The Automated Daily - Hacker News Edition

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily

- Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad

- Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad

Today's topics:

AI finds Linux kernel exploits - OpenAI built an autonomous AI security pipeline that found, reproduced, exploited and patched Linux kernel vulnerabilities. One run produced 948 working privilege escalation exploits for about $1.2 million in API costs.

PHP2Shell and AI-driven security - IEEE Spectrum says the patched PHP2Shell vulnerability shows how AI tools turn old bugs into code execution exploits. The article argues that AI is exposing technical debt in legacy software and calls for AI-assisted auditing and formal verification.

Kimi K3.1 Pro open-weight model - Moonshot AI released Kimi K3.1 Pro, an open-weight multimodal mixture-of-experts model with a 1-million-token context window. It targets agentic coding, long-context reasoning and image understanding.

OpenAI Images 2.6 adds GIFs - OpenAI's Images 2.6 update for ChatGPT, Codex and the API makes image generation faster and editing more consistent. It also adds GIF animation and SVG export.

Go 1.27 ships JSON v2 - Go 1.27 adds the encoding/json/v2 and jsontext standard library packages. They fix long-standing correctness and security problems and add true streaming JSON performance.

Guerrilla bike lanes in San Francisco - Anonymous activists installed protected bike lane posts near Golden Gate Park after two cyclists were killed. San Francisco now plans to make the street safety changes official, a rare win for tactical urbanism.

Why LEDs stopped attracting insects - LED bulbs attract far fewer insects than incandescent lights because they emit little ultraviolet light. That changes how insects navigate, affects predators like geckos and may reduce the spread of some diseases.

-PHP2Shell Shows How AI Is Reshaping Cybersecurity

-Moonshot AI Releases Kimi K3.1 Pro Multimodal Agentic Model

-OpenAI Details Autonomous Linux Vulnerability Discovery Pipeline

-Go Releases New JSON API in the Standard Library

-OpenAI Launches Images 2.6 With Faster Editing and GIFs

-San Francisco Turns Guerrilla Bike Lanes Permanent

-Ricursive Unveils Ripple-65B Preview Open-Weight AI Model

-Why Insects Swarm Less Around LED Lights

Episode Transcript

AI finds Linux kernel exploits
Let's start with that headline. OpenAI says it has built an autonomous security research pipeline that discovers, validates, exploits and even patches Linux kernel vulnerabilities, all inside a controlled environment. In a single campaign, the system flagged more than two thousand candidate issues and reproduced most of them. It produced 948 working local privilege escalation exploits, meaning an unprivileged user could become root. The bill was roughly 1.2 million dollars in API credits. Human researchers reviewed the findings before passing them to kernel maintainers, and the system also suggested candidate fixes. OpenAI's main argument is that finding bugs is no longer the hard part. The real bottlenecks are now deciding which bugs matter most, confirming them and getting patches shipped. The company also says the work turned up new Linux exploitation techniques, which makes a strong case for defense in depth and faster hardening.

PHP2Shell and AI-driven security
That ties directly into a story we've been following. IEEE Spectrum now argues that the PHP2Shell vulnerability is an early warning of how AI is changing cybersecurity. The new detail is how the flaw was found. For years it looked like a harmless out-of-memory crash, until researchers used AI tools to work out how to turn it into arbitrary code execution. The piece warns that AI helps skilled defenders and attackers alike reason through messy legacy code, and that this brings decades of technical debt to the surface. Its recommendations are more funding for defenders, better AI-assisted auditing and stronger formal verification tools. For everyone else, the advice is simple: install security updates quickly.

Kimi K3.1 Pro open-weight model
Over in model land, Moonshot AI has released Kimi K3.1 Pro, an open-weight model that handles both text and images and the first entry in its K3.1 series. It's built for agentic work, meaning tasks where the model plans and takes multiple steps on its own. It's enormous, at nearly three trillion total parameters, although only a small fraction is active at any moment thanks to a mixture-of-experts design. It also offers a one-million-token context window for very long documents and codebases. Moonshot highlights gains in coding, 3D creation, financial research and factual accuracy, and its benchmark tables are competitive with top models. Releasing a model of this size with open weights keeps up the pressure on closed labs.

OpenAI Images 2.6 adds GIFs
OpenAI also shipped Images 2.6 across ChatGPT, Codex and its API. The company says generation is about fifty percent faster and that characters, products and brand details stay more consistent across multiple rounds of edits. The most noticeable addition is GIF animation, which is useful for reaction images, mascots and quick storyboards. Developers get extra controls, including SVG export and a setting that trades scene detail against speed. The broader direction is clear: image tools are moving from one-off pictures toward workflows people can actually use in production.

Go 1.27 ships JSON v2
For the programmers: Go 1.27 now officially includes a new generation of JSON packages in the standard library, after a period as an experiment. The original encoding/json package had well-known problems. It quietly accepted invalid text encoding and duplicate keys, matched field names in surprising case-insensitive ways and hit performance limits. The redesign splits low-level JSON parsing from the higher-level mapping to Go types. It also processes data as a true stream instead of loading whole values into memory first. The API should still feel familiar to Go developers. JSON runs through a huge share of Go software, so stricter defaults and better speed will reach almost the entire ecosystem.

Guerrilla bike lanes in San Francisco
Now to the streets of San Francisco. After two cyclists were killed in hit-and-run crashes near Golden Gate Park, an anonymous group calling itself the San Francisco Transformation Agency installed its own protected bike lane posts. Unauthorized street changes like this are usually removed quickly. This time, the city said it would install official posts and leave the activists' version in place while it reviewed the area. Similar guerrilla efforts in Seattle and New York got mixed reactions. Here, the stunt seems to have sped up changes the city was already open to, given the miles of protected lanes it has built recently.

Why LEDs stopped attracting insects
Finally, a small nature mystery. Have you noticed fewer bugs swarming around porch lights? The likely reason is LEDs. One writer noticed that after a family in Mexico switched to LED bulbs, the insects disappeared from their outdoor dining area, and so did the geckos that used to hunt them. Research suggests insects navigate partly by moonlight cues, and the ultraviolet light from older bulbs confuses them and pulls them toward lamps. LEDs give off little or no UV light, so many flying insects simply ignore them. That could mean less disruption to wildlife and possibly less spread of some insect-borne diseases. LEDs still raise other concerns, though, such as the effect of blue light on circadian rhythms.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- SurveyMonkey, Using AI to surface insights faster and reduce manual analysis time - https://get.surveymonkey.com/tad

- Consensus: AI for Research. Get a free month - https://get.consensus.app/automated_daily

- Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily

Today's topics:

Postman silently installs AI agent plugins - An issue report alleges Postman on macOS auto-installed Codex and Claude Code plugins, changed global agent configs, and self-trusted startup hooks without consent. The case raises developer tool security, MCP, and trust-boundary concerns.

Rooting a Renault to fix clock - A Renault Latitude owner reverse engineered a TomTom Carminat LIVE head unit, chaining a path traversal and Linux core_pattern trick into root access just to fix daylight saving time. The hack exposes weak embedded and car infotainment security.

LLMs reading neural network weights - Simon Willison tested Qwen running locally to see whether an LLM can identify reinforcement learning agents for Pac-Man, Pong, and Breakout from raw weights alone. Results impressed, but he cautions against hype about weight-space programming replacing source code.

C++26 fails ISO DIS vote - The C++26 draft failed the ISO DIS vote for the first time in C++ history, mainly over disagreement about Contracts. Options include revising, revoting, a technical specification, or cancellation, leaving C++26 and C++29 uncertain.

The Internals of PostgreSQL updated - Hironobu Suzuki's The Internals of PostgreSQL now covers PostgreSQL 19 beta, with new material on asynchronous I/O and logical replication. It is a free deep resource on WAL, vacuum, MVCC, and replication.

Safari Adaptive Dark Mode beta - Apple's WebKit team introduced Adaptive Dark Mode in Safari 27.3 beta, automatically converting light websites to dark themes while preserving contrast and readability. Developers are still encouraged to use color-scheme and prefers-color-scheme.

Fujifilm quintuples magnetic tape production - Fujifilm will invest about 10 billion yen to expand LTO magnetic tape production fivefold, driven by generative AI data archiving demand. Tape remains cheap, offline, and a strong defense against ransomware.

-Postman Issue Alleges Silent Plugin Installation and Hook Trust Changes

-Hacking a Renault Navigation Unit to Fix the Car Clock

-Qwen 3.8 27B Can Interpret Model Weights on a Laptop

-Neufangled Unveils Stringman, a Room-Scale Cable Robot

-Fujifilm plans fivefold tape production boost amid AI archiving demand

-WebKit Adds Adaptive Dark Mode to Safari 27.3 Beta

-C++26 DIS Vote Fails Over Contracts, Leaving C++ Future Uncertain

-The Internals of PostgreSQL Homepage and Update Log

Episode Transcript

Postman silently installs AI agent plugins
Let's start with a trust problem in developer tooling. An issue report alleges that simply opening Postman on macOS automatically installed plugins for Codex and Claude Code, rewrote global agent configurations, and added startup hooks that steered general API-related tasks toward Postman, all without an explicit opt-in. According to the reporter, Postman's own logs show an automatic existing-user upgrade doing the installs, after which the installer tried to approve its own hook inside Codex. They also say a remote MCP connection was switched on even though authentication wasn't working, and that a command-line binary landed on the system too. The core complaint is about boundaries: one application shouldn't quietly reconfigure other tools or make itself the default for unrelated work. The reporter is asking Postman to stop the automatic installs, drop the self-approval, clarify consent, and offer a clean rollback. As AI agents gain more power over our machines, who gets to plug into them is becoming a real security question.

Rooting a Renault to fix clock
Now, that car clock. The owner of a 2014 Renault Latitude found that after daylight saving time ended, the built-in TomTom Carminat LIVE unit simply refused to let them adjust the time. The reason turned out to be odd: the time settings only appear when a genuine navigation SD card is inserted, and that card is protected by DRM checks. Rather than pay for a pricey replacement, the author dug into the firmware, built an emulator, and probed the update process and script parser. Several approaches went nowhere until they spotted a path traversal bug in a logging feature. That let them overwrite a Linux kernel setting so that the next program crash would run their code as root. With full control, the fix itself was almost anticlimactic: a single timezone value in a settings file, changed from 180 to 120 minutes. It's a fun read, and also a sober look at how fragile security can be inside car infotainment systems.

LLMs reading neural network weights
An update on the weight-space programming story we covered recently. Simon Willison set aside the bigger claims and asked a narrower question: can a language model tell what a neural network does just by looking at its raw weights, with no tools or code execution? Running a Qwen model locally on his laptop, he found it correctly identified several small reinforcement learning agents trained on games like Pac-Man, Pong, and Breakout. It was slow and its reasoning wandered at length, but the answers were right. It missed on Freeway, likely because that game is more obscure. Willison's takeaway is measured: genuinely intriguing, but not evidence that source code is going obsolete or that AI is ready to meaningfully rewrite itself.

C++26 fails ISO DIS vote
Over in programming languages, a first for C++, and not a welcome one. The C++26 draft failed its ISO DIS vote on October 23rd, which the author of a widely shared post says has never happened before. The sticking point was Contracts, a feature that was pulled from C++20, reworked, and approved several times within the committee, yet several national bodies still judged it not ready. The author argues this creates a procedural and trust problem, since the process depends on consensus and respect for majority decisions. Possible paths forward range from revising and revoting, to a committee review, to spinning Contracts into a technical specification, or even canceling the effort. For now, the timeline for C++26, and by extension C++29, is up in the air.

The Internals of PostgreSQL updated
For database fans, Hironobu Suzuki's long-running guide, The Internals of PostgreSQL, has been updated to cover PostgreSQL 19 beta. It walks through how the database really works under the hood, from query processing and concurrency to vacuuming, write-ahead logging, and replication. Recent additions include material on asynchronous I/O and logical replication. If you've ever wanted to understand why Postgres behaves the way it does, it remains one of the best free resources out there.

Safari Adaptive Dark Mode beta
On the web, Apple's WebKit team announced Adaptive Dark Mode in the Safari 27.3 beta. When enabled, it automatically turns light websites into dark ones, aiming to close the gap between the many people who prefer dark mode and the relatively few sites that support it. Rather than simply flipping colors, it tries to keep text readable and preserve the meaning of things like buttons, icons, and logos. Sites that already offer a dark theme are left alone, and Apple says pages can't detect when the transformation is applied. Still, the company is nudging developers to build proper dark mode support themselves using standard CSS features.

Fujifilm quintuples magnetic tape production
And finally, an old technology getting a fresh boost. Fujifilm plans to spend roughly 10 billion yen to expand its magnetic tape production fivefold, citing demand for long-term archiving fueled by generative AI. The work at its Odawara factory runs through 2028 and focuses on the latest LTO generation. As Ars Technica points out, cloud and data center operators still lean on tape because it's cheap for enormous volumes, and because it sits offline, it's a strong safeguard against ransomware. The AI boom, it turns out, is good news for one of computing's oldest storage formats.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad

- Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad

- Invest Like the Pros with StockMVP - https://www.stock-mvp.com/?via=ron

Today's topics:

AI Edits Neural Network Weights Directly - Early research on Weight-Space Programming reports that GPT-6 Astra can read raw neural network weights, infer what a model was trained to do, fix broken control policies, and write game-playing networks that outperform PPO baselines. The work could open new paths for interpretability, model editing, and automated model design.

Cognitive Surrender to AI Coding Agents - An essay warns that over-reliance on AI agents can lead developers to cognitive surrender, building technical debt and automation bias. It recommends a personal vibe limit, staying in the loop, observability, and simpler modular software design.

Egoless Mathematics in the AI Era - A new essay in the ongoing debate on AI and mathematics argues for egoless mathematics, where truth, understanding, and discovery matter more than authorship, rankings, or who produces the proof.

Archive Websites Before They Vanish - A web archivist urges individuals to save copies of websites they care about using accessible tools and the WARC format, since even long-lived sites and domains can disappear or be repurposed without warning.

Hacker News Front Page Highlights - The November 3, 2026 Hacker News front page features Andrej Karpathy's HackerTube video library, criticism of Ubuntu's handling of a security patch delay, and growing AI influence on PostgreSQL development.

NIST Cybersecurity Guide for Solopreneurs - NIST published a plain-language cybersecurity guide for solopreneurs, covering backups, software updates, and default passwords, with worksheets and use cases for more than 28 million one-person US businesses.

DiscImageStudio Burns Pictures onto Discs - DiscImageStudio is an open-source GPL-3.0 tool that turns images into disc data so burned CDs, DVDs, and Blu-ray discs display visible patterns, with previews, dithering, calibration, and ISO export.

Baby Names Switching Between Genders - An analysis of US Social Security baby-name data from 1936 to 2025 finds 42 names shifted toward girls and just 12 toward boys, with girls' naming trends driving most gender switches.

-LLMs Learn to Read and Edit Neural Network Weights

-Why You Should Back Up Websites You Love

-US Baby Names That Switched Between Boys and Girls

-Hacker Times November 3, 2026 Edition Highlights AI, Security, and Software News

-DiscImageStudio: Optical Disc Image Design and Burning Tool

-NIST Releases Cybersecurity Guide for Solopreneurs

-Resisting Cognitive Surrender

-Egoless Mathematics and AI

-Momentic: AI Testing Platform for Web and Mobile

Episode Transcript

AI Edits Neural Network Weights Directly
Our lead story is a paper on what its authors call Weight-Space Programming. Instead of training a model with gradients and feedback, a frontier language model reads and edits the weights directly. The researchers report that GPT-6 Astra can often tell what a network was built for just by inspecting its parameters. It can recognize which Atari-style game a policy plays, what a diffusion model was trained to generate, and even which language a small transformer learned. It reportedly diagnosed a broken control policy and repaired it by changing only a small fraction of weights, with no access to the environment. The boldest claim is that Astra wrote small game-playing networks from scratch that beat PPO-trained baselines. These are early results and deserve some skepticism, but if they hold up, LLMs could move beyond writing code and prompts into designing and fixing neural networks themselves. That would matter for interpretability as much as for engineering.

Cognitive Surrender to AI Coding Agents
Staying with AI, an essay making the rounds warns about what it calls cognitive surrender: accepting what an AI agent produces without really thinking it through. The author grants that agents are fast and confident, and that teams feel more productive using them. The cost is complexity that quietly builds up until the system is hard to debug or reason about. The proposed fix isn't to stop delegating. It's to be deliberate about it. Set a personal vibe limit for how much complexity you're willing to hand off, and pay attention when you notice yourself clicking through agent output on autopilot. The essay also recommends better observability and validation before handing over fragile work. Its closing point is a familiar one: good design multiplies everything, and simple, modular systems are easier for humans and agents alike.

Egoless Mathematics in the AI Era
The discussion about AI and mathematics that we've been following has a new entry. This essay argues that AI isn't creating a crisis so much as exposing one that already existed: mathematicians had started measuring their worth through output, rankings, and performance. The authors take the view that mathematics is discovered rather than invented, and they propose what they call egoless mathematics, where truth, understanding, and the joy of discovery matter more than who gets credit. Seen that way, a proof found with a machine's help is no less beautiful. The essay suggests that people who embrace AI and people who avoid it can still share the same commitment to truth.

Archive Websites Before They Vanish
Now for a reminder about how fragile the web is. A web archivist describes a fan site they loved that simply disappeared, and whose domain later turned into a host for gambling ads. The lesson is that even sites that have been around for years can vanish without warning. The author's argument is that preservation isn't only a job for libraries and big institutions. Ordinary users can save the pages they care about with easy-to-use tools and standard formats like WARC. In short, if something online matters to you, keep your own copy.

Hacker News Front Page Highlights
Today's Hacker News front page also includes Andrej Karpathy's HackerTube, which collects two decades of video recommendations from the community into a browsable library. There's also pointed criticism of how Ubuntu handled a reported delay in a security patch, and a discussion of AI's growing role in PostgreSQL development. AI and security continue to dominate the conversation.

NIST Cybersecurity Guide for Solopreneurs
On the practical security side, NIST has published a cybersecurity guide written for solopreneurs, meaning businesses where the owner is the only paid employee. It updates an earlier small-business publication and is written in plain language. It sticks to basics like backing up data, keeping software up to date, and changing default passwords, and it includes worksheets and fictional examples. More than 28 million US businesses fit that description, so simple advice like this could reach a lot of people.

DiscImageStudio Burns Pictures onto Discs
For something more playful, DiscImageStudio is an open-source tool that lets you burn a picture onto a disc. It converts an image into disc data so the surface of a burned CD, DVD, or Blu-ray shows a visible pattern. It offers layering, previews, and a calibration workflow to make the physical result more accurate. It's a clever use of old media, and it's free under the GPL.

Baby Names Switching Between Genders
Finally, an analysis of US baby names from 1936 to 2025 looks at names that switched between boys and girls. Forty-two names moved toward girls, compared with only twelve that moved toward boys. Girls' naming trends drove the change in both directions. Names became more feminine when more girls started receiving them, and names became more masculine mostly because girls stopped getting them. Many names that shifted toward girls did so at their peak popularity, while those that shifted toward boys often switched only after they'd already faded.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- Lindy is your ultimate AI assistant that proactively manages your inbox - https://try.lindy.ai/tad

- SurveyMonkey, Using AI to surface insights faster and reduce manual analysis time - https://get.surveymonkey.com/tad

- Consensus: AI for Research. Get a free month - https://get.consensus.app/automated_daily

Today's topics:

Lenovo fingerprint drive skips encryption - A security researcher found Lenovo's ThinkPlus FU100 fingerprint USB drive stores data unencrypted on a plain SD card. The fingerprint sensor only acts as a switch, and related secure USB products had backdoor accounts and hardcoded keys.

PostgreSQL 19 delayed by AI-found bugs - The PostgreSQL 19 release is slipping toward November after late feature reverts. A surge in LLM-assisted bug reports has sharply increased back-patched fixes and stretched maintainer time.

Benchmarking JavaScript frameworks for AI agents - Nolan Lawson benchmarked coding agents building the same app across JavaScript frameworks. Lighter options like Hono with htmx and Astro were cheaper and faster than SvelteKit, Lit, and Angular, and code size now drives token costs.

Terence Tao on AI and mathematics - Terence Tao argues AI has fundamentally changed mathematics as AI-assisted proofs become easier. He says mathematicians should value problem selection, explanation, and research direction, and should reform hiring, journals, and graduate training.

OccupySaaS clean-room open-source rebuilds - OccupySaaS is a protest-style open-source movement in which laid-off engineers build clean-room, drop-in replacements for SaaS products. Its first releases are Apache-2.0 alternatives compatible with Auth0 and Salesforce.

OmniDiff content-aware open-source diffing - OmniDiff is an open-source diff tool that compares code, images, PDFs, archives, and binary files with type-aware methods. It integrates with Git and Jujutsu and publishes transparent benchmark data.

Hacker Books tracks HN reading favorites - Hacker Books ranks the books most mentioned and recommended in Hacker News comments. Recurring favorites include Gödel, Escher, Bach, Thinking, Fast and Slow, and The Mythical Man-Month.

Outcrop brings geologic maps to AR - Outcrop is an offline geology app that overlays geologic maps on your phone's camera view. It draws on Macrostrat, USGS, and Paleobiology Database data to show rock ages, faults, and fossils.

-Hacker Books Ranks Most-Mentioned Books from Hacker News

-PostgreSQL 19 Delayed as AI Finds More Bugs

-Lenovo Fingerprint USB Drive Found to Be Unencrypted

-Benchmarking the AI Cost of JavaScript Frameworks

-OccupySaaS Launches Clean-Room Open-Source Replacements for SaaS

-Terry Tao Guest Post Argues AI Is Reshaping Mathematics

-OmniDiff Launches as a Fast, Content-Aware Diff Tool

-Outcrop Brings Geologic Maps Into Augmented Reality

Episode Transcript

Lenovo fingerprint drive skips encryption
Let's begin with that hardware story. A security researcher took apart Lenovo's ThinkPlus FU100, a USB drive sold with fingerprint protection, and found that it doesn't encrypt anything. The sensor only enables the USB connection. The storage inside is a plain SD card, so anyone who has the device in hand can pull the card and read the data. The researcher also turned up more serious flaws in other supposedly secure USB products, including backdoor accounts and hardcoded keys. Lenovo had already shipped fixes for several related devices, but some of those fixes require customers to contact support or send hardware back. The lesson is an old one. A security claim on the box tells you nothing about how the device actually works.

PostgreSQL 19 delayed by AI-found bugs
Next, the story of how AI is reshaping software quality. PostgreSQL 19 looks set to arrive later than usual, likely in November or even later, after several major features were pulled late in the beta cycle. A post analyzing the delay blames a flood of bug reports found with the help of large language models. Commit activity on the main branch barely changed. Back-patched fixes to older, supported releases, however, jumped sharply, which means maintainers are spending much more time on maintenance across the board. The author doesn't see this as a crisis. If anything, it shows PostgreSQL's quality bar is holding. The broader point is that bugs which stayed hidden for years are now much easier to find, and every software project should expect that kind of scrutiny.

Benchmarking JavaScript frameworks for AI agents
Staying with AI-assisted development, Nolan Lawson built a benchmark to measure what frameworks cost when coding agents do the work. He had agents build the same small Pokémon app in several JavaScript frameworks, then compared cost, build time, token usage, performance, accessibility, and code size. Lighter, server-oriented approaches such as Hono with htmx, and Astro, generally came out cheaper and faster than heavier options like SvelteKit, Lit-based setups, and Angular. One conclusion stands out: code size matters more than ever, because every extra line becomes tokens the agent has to read and revise later. Lawson stresses that the benchmark is imperfect and not a final verdict on any framework. His argument is that these choices should rest on data rather than gut feeling, and he invites others to reproduce or extend the work.

Terence Tao on AI and mathematics
From code to proofs. Terence Tao has published an essay arguing that AI has fundamentally changed mathematics, and that mathematicians should adapt to it rather than grieve. As AI makes new proofs easier to produce, he argues, the human work that matters more is choosing worthwhile problems, explaining why results are important, and opening new lines of research. He pushes back on the idea that AI will 'solve' a field, noting that past breakthroughs reshaped mathematics without exhausting it. He also warns the profession against rewarding speed and paper counts over real understanding. His proposals include judging work as if it were AI-assisted, placing more weight on in-person hiring, asking more of journals, and broadening graduate training. It's a call to raise ambitions, not lower standards.

OccupySaaS clean-room open-source rebuilds
Turning to open source, a movement called OccupySaaS launched last month with a provocative premise: laid-off engineers rebuilding popular SaaS products as clean-room, drop-in compatible replacements. The group criticizes per-seat pricing and vendor lock-in. It says its projects rely only on public documentation, open standards, and compatibility testing, with no leaked code. Its first releases are a self-hostable identity service compatible with Auth0 and a CRM compatible with Salesforce, both under the Apache license. The group describes a strict process in which spec writers and builders work separately and the origin of all code is tracked. Whether it gains real traction or not, it reflects how much frustration exists over layoffs and cloud software costs.

OmniDiff content-aware open-source diffing
Also in open source, a new project called OmniDiff aims to replace the plain-text diff. It uses comparison methods suited to each file type, so it can handle images, PDFs, fonts, archives, and other binary files as well as code. It works with Git, Jujutsu, editors, and CI pipelines. What's notable is how much benchmark and quality data the project publishes, which matters for teams whose repositories contain far more than source code.

Hacker Books tracks HN reading favorites
A couple of lighter discoveries to finish. Hacker Books collects the books mentioned in Hacker News comments and ranks them by how often they're cited and recommended. The usual favorites show up near the top, among them Gödel, Escher, Bach, Thinking, Fast and Slow, The Mythical Man-Month, and How to Win Friends and Influence People. It turns years of scattered recommendations into a reading list, and it gives a fair picture of what tech culture keeps coming back to.

Outcrop brings geologic maps to AR
And for anyone heading outdoors, Outcrop is a geology app that overlays geologic maps on your phone's camera view, even when you're offline. Point it at a hillside and it tells you what rock you're looking at, how old it is, and where on the globe it formed. It also shows faults and fossils, using open data from Macrostrat, the USGS, and the Paleobiology Database. It's a handy way to turn an ordinary hike into a lesson in the landscape's history.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- Lindy is your ultimate AI assistant that proactively manages your inbox - https://try.lindy.ai/tad

- Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad

- KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad

Today's topics:

Karpathy builds HackerTube with Claude - Andrej Karpathy used Claude Opus 5.5 to build HackerTube, a YouTube-style site of Hacker News-recommended videos, in about two hours. It shows how AI coding agents can rapidly ship niche products.

ChatGPT and Codex share usage limits - OpenAI will share usage allowances across ChatGPT and Codex for Plus, Pro, and Business plans starting in December. Usage is measured in tokens, and instant chat does not count against the limit.

ChatGPT shared-link account takeover bug - Researchers at Parameter found a ChatGPT vulnerability that allowed account takeover through a shared conversation link by chaining HTML Preview, MathLive, and a CSP bypass. OpenAI fixed it and paid a $5,000 bug bounty.

Critical Linux TCP kernel vulnerability - CVE-2026-98432 is a critical remote use-after-free in the Linux IPv4 TCP handshake affecting kernels back to 4.4. Multipath TCP listeners, including Go defaults, raise the risk, and some Ubuntu and RHEL kernels were still unpatched.

Jepsen exposes ParticleDB correctness failures - Jepsen testing of ParticleDB found 96 issues, including data loss, lost updates, and corruption, despite Full ACID claims. Many failures occurred on a single healthy node without injected faults.

Why GROUP BY queries stall on memory - A database performance deep dive shows that hash-based GROUP BY aggregation is often limited by memory latency rather than CPU. Radix partitioning cut DRAM traffic and cache misses, making queries faster despite extra work.

Four-neuron sparks show emergent behavior - An interactive neuroscience explainer builds tiny creatures called sparks from just a few neurons. Changing wiring and connection strengths produces attraction, fear, aggression, and love-like behavior.

Italian farmer revives ancient wheat - Italian farmer Claudio Grossi planted forgotten attic seeds and revived poulard wheat, an ancient Triticum turgidum variety. Its hardiness, minerals, and low-input needs make heritage grains relevant amid climate change.

Norway's architectural scenic route toilets - Norway's National Tourist Routes feature public toilets designed by leading architects at scenic spots like Stegastein and Ureddplassen, turning basic infrastructure into tourist attractions.

-Karpathy Builds “HackerTube” as an AI-Powered YouTube Alternative

-Critical Linux TCP Use-After-Free Exposes Servers to Remote Attack

-Italian Farmer’s Attic Seeds Turn Out to Be Ancient Miracle Wheat

-Researchers Say ChatGPT Bug Enabled Account Takeover via Shared Link

-Jepsen Finds Serious Correctness and Durability Bugs in ParticleDB

-Memdoor team-solve workflow reproduces a multi-agent coding benchmark

-How a Tiny Brain Creates Complex Behavior

-Why Cache-Friendly Partitioning Speeds Up GROUP BY

-Norway’s Scenic Routes Feature World-Famous Designer Toilets

-OpenAI to Share ChatGPT and Codex Usage Limits Across Plans

Episode Transcript

Karpathy builds HackerTube with Claude
Let's start with Andrej Karpathy, who got fed up with YouTube's increasingly noisy, clickbait-driven recommendations and decided to build his own alternative. He asked Claude Opus 5.5 to make a YouTube-style site fed entirely by videos people have recommended on Hacker News. About two hours later, he had HackerTube. The system combed through nearly two decades of HN comments, pulled out video recommendations, and organized more than six thousand videos into roughly 150 topics. It offers recommendations without accounts and lets anyone export the data. The more striking part is Karpathy's follow-up: he says Claude Code now handles most of his engineering, and his own job has shrunk to things like getting an API key, buying a domain, and spinning up a server. Whether that sounds exciting or unsettling probably depends on where you sit, but it's a clear snapshot of where AI-assisted development is heading.

ChatGPT and Codex share usage limits
Staying with AI, OpenAI announced that starting in December, usage allowances on Plus, Pro, and Business plans will be pooled across ChatGPT and Codex. Consumption will be counted in tokens, so heavier models and harder tasks burn through your allowance faster, while instant chat won't count at all. OpenAI says most people shouldn't hit resets more often than they do today, and you'll be able to track your combined usage in settings. For developers who lean on Codex heavily, it's worth keeping an eye on.

ChatGPT shared-link account takeover bug
Now to that hook. Security researchers at Parameter found a serious flaw in ChatGPT that could let an attacker hijack another user's account simply by getting them to open a shared conversation link. The attack stitched together several features that each seemed harmless on their own, including the sandboxed HTML preview and a graphing component, until untrusted code slipped into the main ChatGPT page and ran as the victim. From there it could read private chats and make requests on their behalf. The researchers even sidestepped the site's content security policy using an allowed YouTube endpoint. OpenAI patched it on October 26th, rated it high priority, and paid a five thousand dollar bounty. The lesson is a familiar one: reasonable features can combine into something dangerous when trust boundaries get fuzzy.

Critical Linux TCP kernel vulnerability
On the infrastructure side, Linux researchers disclosed a critical vulnerability in how the kernel handles the IPv4 TCP handshake, affecting kernels as far back as version 4.4. Because it triggers before any application-level authentication, a remote attacker with no credentials could potentially leak memory, crash a machine, or even run code. Servers using Multipath TCP are especially exposed, and notably, Go programs may enable that by default on Linux. The upstream fix landed in October, but some major distribution kernels, including Ubuntu and Red Hat Enterprise Linux, were reportedly still unpatched. Web servers, SSH hosts, databases, and Kubernetes nodes are all on the list, so this is one to patch promptly.

Jepsen exposes ParticleDB correctness failures
Next, a sobering report from Jepsen, which put ParticleDB through its paces. The database is marketed as a fast, fully ACID-compliant option for AI workloads, but Jepsen found close to a hundred issues, from crashes and deadlocks to lost data and genuinely odd corruption, like rows that refused to be deleted and tables that came back in mangled form. What makes it worse is that most of these problems showed up on a single, healthy node with no faults injected. Even the latest release candidate could still lose committed writes by default. When a database's core promise is keeping your data correct, that's a big gap between marketing and reality.

Why GROUP BY queries stall on memory
Sticking with databases, a nice technical write-up looks at why GROUP BY queries can feel slow even when your CPU appears fully busy. The answer is that the processor is often just waiting on memory. The author shows that by splitting the data into many smaller chunks up front, each piece fits better in the CPU's cache, which cuts memory traffic dramatically. It means doing more work overall, yet finishing faster. A good reminder that in performance tuning, the obvious bottleneck isn't always the real one.

Four-neuron sparks show emergent behavior
Shifting to something more whimsical, an interactive article builds tiny creatures called sparks from a handful of neurons. With one eye and one flipper, a spark can react to light but can't steer. Add a second pair, cross the wiring, and change signals from excitatory to inhibitory, and suddenly you get behavior that looks like attraction, fear, aggression, or even affection, all from four neurons and two connections. Tweak the connection strengths and individual sparks develop their own preferences. It's a charming, approachable look at how behavior can emerge from simple wiring.

Italian farmer revives ancient wheat
From brains to bread. In the early 1990s, Italian farmer Claudio Grossi found an old sack of seeds in his family attic and planted them out of curiosity. The plants grew unusually tall, with seven ears of grain per stem, which reminded him of a passage in Genesis. Testing later identified it as poulard wheat, an ancient variety his grandfather had likely saved. It turned out to be richer in minerals, lower in gluten, and remarkably hardy, needing little in the way of inputs. With drought and soil stress pressuring modern farming, Grossi's so-called miracle wheat has become part of a wider revival of heritage grains.

Norway's architectural scenic route toilets
And finally, Norway has turned public restrooms into destinations. Along its National Tourist Routes, toilets designed by notable architecture firms sit at spectacular viewpoints overlooking fjords, waterfalls, and even the northern lights. Some are tied to wartime memorials, and many are remote and only open seasonally. It's a lighthearted story, but it makes a real point: even the most basic infrastructure can be thoughtfully designed, and become part of the attraction.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad

- Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad

- Invest Like the Pros with StockMVP - https://www.stock-mvp.com/?via=ron

Today's topics:

Binance $6.3 Billion Crypto Heist - Attackers drained about $6.3 billion from Binance wallets across several blockchains in roughly 20 minutes. Tracing firms suspect North Korea's TraderTraitor group, and only about half the stolen crypto has been frozen.

Claude Code Extension SSH Backdoor - Phorion found that the Claude Code for VS Code extension on the Cursor Marketplace was compromised with a persistent SSH backdoor. The malware added attacker SSH keys, installed persistence on macOS, Linux and Windows, and sent reconnaissance beacons from developer machines.

Northflank Security Incident and Outage - Northflank reported a partial outage tied to a security incident involving unauthorized use of customer AI API keys. The platform told users to assume their environment variables were decrypted and to rotate all secrets and credentials.

Azure API Connections Cross-Tenant Flaws - A researcher disclosed five critical Azure Logic Apps and API Connections vulnerabilities that allowed cross-tenant access to Key Vaults and SQL databases. Path traversal and path-normalization bypasses defeated several of Microsoft's patches in a row.

Anti-Bot Laws Versus Web Scraping - Proposed laws such as the Stealth Bot Prohibition Act aim to curb abusive AI crawlers. Critics argue the identity-disclosure rules would hurt journalists, researchers, archivists and accessibility auditors who rely on anonymous web scraping.

Cloudflare Turnstile Solver Without Browsers - cfsolve is an open-source Rust library and CLI that solves Cloudflare Turnstile challenges using browser impersonation instead of a headless browser. It highlights the ongoing arms race between bot protection and automated access.

Classic Mac OS on Nintendo Wii - A hobbyist ported classic Mac OS to Nintendo GameCube and Wii hardware, booting Mac OS 7.6 and 9.2.2 through a patched OpenBIOS and custom drivers. The project shows the deep similarities between Nintendo's PowerPC consoles and vintage Macs.

Nesformer Update: NES Inside Qwen - Nesformer compiles an NES 6502 CPU, its RAM and a cartridge ROM into the weights of a stock Qwen3.5 language model. Building on Qwenputer, it runs test ROMs frame-for-frame identically to a conventional emulator.

Florilegium Procedural LaTeX Floral Margins - Florilegium is a LuaLaTeX package that procedurally draws Delft-blue floral page margins at PDF build time. It supports seasonal themes, decorative initials and plants that grow to fill the available space on the page.

-Phorion Warns of Backdoored Claude Code VS Code Extension

-Binance Hit by a $6.3 Billion Crypto Heist

-Northflank Reports Security Incident and Major Platform Outages

-Classic Mac OS Reaches the Nintendo Wii

-Five Critical Azure API Connection Flaws Enable Cross-Tenant Compromise ([binarysecurity.no](https://www.binarysecurity.no/posts/2026/10/one-root-case))

-Tangled Launches a Federated Code Forge

-Florilegium: Procedural Floral Margins for LuaLaTeX

-Why Anti-Bot Laws Could Hurt Legitimate Web Crawling

-NES CPU Compiled Into Qwen3.5 Model Weights

-Rust Tool for Solving Cloudflare Turnstile and Challenge Pages

Episode Transcript

Binance $6.3 Billion Crypto Heist
Let's begin with that Binance story. Crypto firms and investigators report that attackers drained about six point three billion dollars from Binance wallets spread across multiple blockchains, all in around twenty minutes. Binance quickly froze deposits and withdrawals while it traced the breach. Several tracing firms point toward North Korea's TraderTraitor group, though neither Binance nor law enforcement has confirmed that. What stands out is the preparation: test transactions and apparent rehearsals before the main event. Roughly half the funds have reportedly been frozen, but the rest sits in assets that can't simply be locked. It's a stark reminder that once money starts hopping between chains, defenders are mostly playing catch-up.

Claude Code Extension SSH Backdoor
Staying on security, but closer to every developer's laptop. Phorion reports that the official Claude Code for VS Code extension on the Cursor Marketplace was compromised and shipped with a persistent SSH backdoor. As soon as the extension loaded, it planted an attacker's key in the user's authorized SSH keys, tried to open an outbound connection to an attacker server, and set up persistence on macOS, Linux and Windows so it would survive reboots. It also sent home a beacon describing the machine and hinting at what secrets might be on it. The two affected versions have been pulled, but the lesson is uncomfortable: a trusted coding tool is about the most direct route into source code and internal credentials an attacker could ask for. If you used it, check your authorized keys.

Northflank Security Incident and Outage
Another incident is unfolding at the cloud platform Northflank. Its status page lists a partial outage linked to a security incident, after users reported unauthorized use of AI API keys they had stored on the platform. The company's guidance is blunt: assume your environment variables were decrypted and rotate every secret and credential right away. Networking, builds, services and logs are among the components marked in major outage. Northflank has turned on audit logs for all customers, including free users, and says the investigation continues.

Azure API Connections Cross-Tenant Flaws
And on the cloud provider side, a researcher has detailed five critical vulnerabilities in Azure's API Connections, the plumbing behind Logic Apps. The core problem: Azure could be tricked into using a victim's connection credentials instead of the attacker's, opening the door to another tenant's Key Vaults and SQL databases. What makes the write-up memorable is the cat-and-mouse with Microsoft. Each fix closed one route, the researcher found another, and finally a path-normalization quirk slipped right past the newest validation. It's a case study in how patching symptoms can leave a fragile design underneath.

Anti-Bot Laws Versus Web Scraping
Shifting to policy. A new article takes aim at proposed anti-bot laws, including the federal Stealth Bot Prohibition Act and a similar New York bill. The authors agree aggressive AI crawling is a real problem, but argue that forcing crawlers to identify themselves would mostly hurt the good actors: investigative journalists, the Internet Archive, accessibility auditors, public health researchers and security teams. Their point is simple: knowing who a crawler is doesn't tell you whether it's causing harm. Instead, they want rules targeting concrete damage, like excessive request volumes or fraud, plus better technical standards for machine access.

Cloudflare Turnstile Solver Without Browsers
That debate gets a practical backdrop from a new open-source Rust project called cfsolve, which solves Cloudflare Turnstile challenges without spinning up a headless browser, relying instead on impersonating specific browser profiles. Whatever your view on it, tools like this show why the bot-blocking arms race is so hard to win with technology or legislation alone.

Classic Mac OS on Nintendo Wii
Now for the fun corner. A hobbyist has gotten classic Mac OS running on Nintendo GameCube and Wii hardware, with Wii U on the wish list. Getting there meant porting OpenBIOS, rewriting assumptions baked into Apple's software, and building custom drivers for disks, graphics and controllers. After plenty of false starts, the screen went from a blinking floppy to the Happy Mac, and eventually to working desktops in Mac OS 7.6 and 9.2.2. It's a lovely reminder of how closely related those PowerPC consoles are to vintage Macs.

Nesformer Update: NES Inside Qwen
An update on a story we've followed: the Qwenputer idea has grown a sequel. A project called nesformer now compiles an entire NES CPU, its RAM and a game cartridge into the weights of a stock Qwen language model, with a small host program handling graphics and input. The headline result is that bundled test ROMs match a conventional emulator frame for frame. Odd, yes, but a genuinely striking demonstration of what can be packed into a neural network.

Florilegium Procedural LaTeX Floral Margins
Finally, something for the typography fans. Florilegium is a LuaLaTeX package that draws Delft-blue flowers in your document margins, generating every stem and petal in code when the PDF is built, no images involved. It can grow a plant to fill empty space on the last page, tuck blooms beside headings, and shift the palette from spring to autumn. Every page ends up feeling hand-illustrated.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad

- Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad

- SurveyMonkey, Using AI to surface insights faster and reduce manual analysis time - https://get.surveymonkey.com/tad

Today's topics:

Deadly High-Speed Train Crash in Spain - A collision between Renfe and Iryo high-speed trains near Toledo, Spain, killed at least 30 people. A computer failure and the decision to resume service without speed restrictions are under scrutiny.

Mistral Large 4 Open Weights - Mistral released open weights for Mistral Large 4, a multimodal model for coding, cybersecurity, and enterprise work. The company is promoting it as a self-hostable, European-trained option for sovereign AI.

Fifty AI Agents Build Minecraft Civilization - Y Combinator Labs ran Civ50, an experiment in which 50 GPT-6.1 agents tried to build a civilization in Minecraft. The agents formed a hierarchy, shared plans, and rationed food after many of them died on the first night.

Tiny World Model on Raspberry Pi - A hobbyist project runs a looped transformer world model and a reinforcement learning agent on a Raspberry Pi Pico 2 W. The system can dream up and play simple games such as Pong, Breakout, and a Doom-lite on a microcontroller.

Martin Fowler on LLM Voice - Martin Fowler says readers can often sense the LLM voice in AI-drafted prose, even after heavy editing. He argues that writing is a thinking tool and that AI writing help should be used selectively and transparently.

How H.264 Compresses Video - An explainer shows why H.264 video can be smaller than a single PNG screenshot. The codec combines lossy frequency compression, chroma subsampling, motion compensation, and entropy coding.

World Map Redrawn by IPv4 - An interactive map resizes each country according to its share of the global IPv4 address pool. It shows how unevenly internet addresses are distributed around the world.

-Mistral releases open-weight Mistral Large 4 for coding and cybersecurity

-Arkham Profile for Binance-Hacker-1029

-Martin Fowler Warns Against the “LLM Voice” in Writing

-How H.264 Compresses Video So Efficiently

-YC Labs Tests Multi-Agent Swarms in Minecraft

-If IPv4 Were Land: A World Map Redrawn by IP Addresses

-Herdr: Desktop App for Terminals and Coding Agents

-Tiny World Model Runs and Plays Games on a Pico 2 W

-Deadly Spain Train Crash Near Madrid Leaves Dozens Dead

Episode Transcript

Deadly High-Speed Train Crash in Spain
We start with sad news from Spain. A high-speed train run by Renfe collided with an Iryo train on the Madrid to Seville route near Toledo, south of the capital. At least 30 people were killed and many more were injured. The two trains were carrying more than 500 passengers combined. Several carriages derailed, and some fell down an embankment. Officials say a computer failure first stopped several trains on the line. The transport minister has suggested that letting those trains move again without speed restrictions may have contributed to the crash. Authorities expect the death toll to rise as rescuers keep searching the wreckage. It is already one of the worst rail disasters in modern European history. The investigation will focus on control-room procedures and on whether human error, technical failure, or both led to the collision.

Mistral Large 4 Open Weights
In AI news, Mistral has released the weights for Mistral Large 4. It's an open-weight, natively multimodal model aimed at coding, cybersecurity, and general enterprise work. Compared with the preview release, Mistral says the final model got more reinforcement learning training. The company reports better coding results and lower token usage. It also points to strong scores on security benchmarks and a high rate of refusing malicious hacking requests. Users can switch between three reasoning modes to balance cost against capability. The bigger message is sovereign AI. The model was trained in Europe, and organizations can inspect it, customize it, and run it on their own hardware. That pitch appeals to companies and governments that are uneasy about sending sensitive data to someone else's servers.

Fifty AI Agents Build Minecraft Civilization
Back to those Minecraft agents. Y Combinator Labs built a framework that lets long-running AI agents observe the game, remember what happened, talk to each other, and run scripts. Their headline experiment, called Civ50, gave 50 GPT-6.1 agents three hours to build a thriving civilization. The project turned into a survival and logistics problem almost immediately. The agents cut down every nearby tree, organized themselves into a hierarchy, wrote shared plans, and set up a food rationing system. The first night was brutal. So many agents died that the survivors became noticeably more cautious and focused on building shelter. The researchers say the goal is to study how large groups of agents settle into roles and develop forms of governance under pressure. Going by these results, they come up with something like rationing and a chain of command.

Tiny World Model on Raspberry Pi
Staying with small-scale AI, one developer has put a world model on a Raspberry Pi Pico 2 W, a microcontroller that costs a few dollars. One small neural network predicts the next frame of a game and its reward. A second network plays inside that dreamed-up world and learns from it. The repository covers simple games including Pong, Breakout, Flappy Bird, a Pac-Man lite, and a Doom-lite. The model reproduces most frames exactly, though Doom-lite and Pac-Man are harder for it. The author also checks the agent's dreamed transitions against the real game rules, rather than only against recorded training data. The result shows how much a very small model can do on cheap embedded hardware.

Martin Fowler on LLM Voice
Martin Fowler, the well-known software writer, has published a short essay on what he calls the LLM voice. His point is that readers can often tell when prose was drafted by a language model, even if they can't say exactly why. He compares the reaction to the uncanny valley, and he doubts detection tools can reliably settle the question. Fowler warns that even heavy human editing may not remove that artificial feel. He also makes a broader argument that writing is how many of us think, so handing it off can weaken our reasoning. He sees outlining, translation, and accessibility as more defensible uses. One example he gives is helping someone with Parkinson's disease write. He also accepts AI for routine documents that nobody will read closely. His overall advice is to use AI selectively and openly while keeping your own voice.

How H.264 Compresses Video
For something more hands-on, a popular explainer looks at why a long H.264 video can be smaller than a single PNG screenshot of the same webpage. The short answer is that the codec stacks several tricks. It drops fine detail the eye barely notices. It stores color at lower resolution than brightness, because we're more sensitive to brightness. Then it saves one full frame and describes the following frames mostly as motion relative to it, which is why mostly static footage compresses so well. A final coding pass removes whatever redundancy is left. The author's conclusion is that none of this is magic, just decades of careful engineering.

World Map Redrawn by IPv4
Finally, an interactive visualization redraws the world map by sizing each country according to its share of the global IPv4 address pool instead of its land area. The math compares about 135 million square kilometers of land outside Antarctica with roughly 3.6 billion locatable addresses. Some countries swell dramatically and others nearly disappear. The map turns dry allocation tables into an immediate picture of how concentrated internet address ownership is.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad

- Invest Like the Pros with StockMVP - https://www.stock-mvp.com/?via=ron

- Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily

Today's topics:

Anti-LLM obfuscation in fingerprinting scripts - A crawl of 10,000 websites for canvas fingerprinting led to a Servicepipe anti-bot script stuffed with bioweapon-themed names, likely meant to make AI deobfuscation tools refuse to analyze it. Underneath, it collects ordinary browser and WebGL signals.

Apple still trusts compromised eMudhra certificates - Eleven days after Apple was notified, macOS developer tools like git, curl and python3 still trust the compromised eMudhra root certificates. macOS 27.2 even undid a manual workaround, raising man-in-the-middle and software supply chain risks.

Google unveils Gemini 4 Carbon - Google announced Gemini 4 Carbon, its strongest frontier model, which tops the Artificial Analysis Intelligence Index and shows big gains in coding, agentic workflows and cybersecurity benchmarks. Access starts with cyber defenders through the Fairwind Program.

Qwenputer: a CPU inside model weights - Qwenputer embeds a deterministic RISC-V CPU into Qwen 3.5 4B weights, using Gated DeltaNet layers as exact RAM so code runs inside forward passes. It runs C and MicroPython and adds an in-model persistent filesystem.

Claude Code's hidden summary token costs - A gist claims Claude Code's internal agent_summary status updates for background agents consume roughly 15 to 20 percent of tokens or API requests. A Bun preload script suppresses the requests to save token budget and rate limits.

Carmelite monks carve Gothic stone with CNC - The Carmelite Monks use CNC machines with diamond tools and 3D modeling to carve detailed Gothic stonework for their church. Modern automation makes traditional Gothic craftsmanship feasible again.

Cancer recovery collides with PHP emergency - A developer chronicles a year of lymphoma treatment, chemotherapy, pneumonia and recovery, including a late-night hospital bed scramble to secure company servers against a newly disclosed PHP vulnerability.

-Fingerprinting Script Hides Ordinary Anti-Bot Checks Behind Bioweapon References

-Apple’s macOS Developer Tools Still Trust Compromised eMudhra Certificates

-Google Announces Gemini 4 Carbon Frontier Model

-Apple unveils MacBook Pro with PROLED XDR display

-Qwenputer: a RISC-V CPU compiled into Qwen 3.5

-CircuitHub Unveils The Grid for Fast U.S. PCB Manufacturing

-How Carmelite Monks Use CNC Machines to Carve Gothic Stone

-Workaround Claims Claude Code Wastes Tokens on Background Agent Summaries

-A Year of Cancer Treatment, Recovery, and a Late-Night Server Crisis

Episode Transcript

Anti-LLM obfuscation in fingerprinting scripts
Let's start with that hook. A researcher crawled ten thousand websites to collect canvas fingerprinting and anti-bot scripts, then turned the results into a public gallery at fingerprint.garden. One unusually artistic canvas led back to a script hosted on servicepipe.tech, and inside its obfuscated code were bizarre variable names and phrases about bioweapons and other dangerous subjects. The author's theory: this is anti-LLM obfuscation, designed to trip the safety filters of AI assistants so they refuse to help deobfuscate the code. Once untangled, though, the payload was pretty mundane, gathering the usual browser and graphics signals like user agent, device memory, language and GPU details. Domain digging tied it to Servicepipe, a Russian anti-DDoS and traffic-control company. The takeaway is that fingerprinting vendors may now be designing code to frustrate both human reverse engineers and the AI tools they lean on.

Apple still trusts compromised eMudhra certificates
Staying with security, an update on the eMudhra certificate incident. Apple was notified on October 17th, yet as of October 28th, macOS developer tools including git, curl and python3 still trusted the compromised certificates. That leaves room for attackers to intercept HTTPS traffic, tamper with Git clones and pulls, or lift credentials like GitHub tokens. Worse, macOS 27.2, released on October 26th, didn't fix the problem and actually reverted a manual workaround some users had applied. Plenty of other projects, including volunteer-run ones, have already shipped fixes. Developers are advised to recheck and reapply the workaround after every update, and security teams should check whether any Macs sit in their CI or build pipelines, since this is a supply chain concern, not just a laptop one.

Google unveils Gemini 4 Carbon
On to AI. Google has announced Gemini 4 Carbon, calling it its strongest model yet. Google says it ranks first on the updated Artificial Analysis Intelligence Index and makes big jumps in coding, agentic work and security benchmarks, with the ability to keep chipping away at hard bugs in large codebases for days. The company also highlights better resistance to prompt injection, fewer unsupported claims and less cheating in internal tests. Notably, the rollout starts small, with a limited group of cyber defenders through what it calls the Fairwind Program, before reaching developers and consumers. It's a clear signal that frontier labs see software engineering and cybersecurity as the next big battleground, and are trying to pair raw capability with tighter controls.

Qwenputer: a CPU inside model weights
From the giant to the genuinely strange. A project called Qwenputer takes the small Qwen 3.5 4B model and compiles a working RISC-V CPU directly into its weights. Certain layers of the model act as exact, addressable memory, so programs execute entirely inside the model's normal forward passes, with no external sandbox or tools. The author reports it passes standard RISC-V tests, matches the QEMU emulator, and has run millions of instructions without errors, handling arithmetic, C and MicroPython while leaving the model's usual behavior nearly untouched. There's even an experimental in-model filesystem that persists files across chats. Beyond the novelty, it hints at ways to bake reliable computation and memory into models, with possible uses in AI safety and monitoring research.

Claude Code's hidden summary token costs
And if you use Claude Code, this one might touch your wallet. A widely shared gist argues that a background feature called agent summary, which asks a model to describe a background agent's recent work in a few words, quietly eats around fifteen to twenty percent of spend or API requests. According to the author, it only updates a status line in the agents panel and has no effect on the actual conversation. The gist offers a preload script that suppresses those requests without modifying Claude's binary, with the caveat that it could silently stop working after updates. Whether or not the exact numbers hold for everyone, it's a good reminder that agentic tools can carry hidden overhead worth auditing.

Carmelite monks carve Gothic stone with CNC
Now for something a little more timeless. The Carmelite Monks, who are building a Gothic church, explained how they use CNC machines to carve stone. Their argument is that modern machinery isn't out of place in medieval-style architecture; it's what makes richly detailed Gothic work practical again. An artist builds 3D models, programmers turn them into toolpaths, and diamond-tipped machines can run overnight while sawyers prepare the stone. The result is ornate carving that would once have taken years by hand, in service of a building meant to last for centuries. Old craft, new tools, same ambition.

Cancer recovery collides with PHP emergency
Finally, a personal post that resonated with many readers. A developer shared his year of lymphoma treatment: painful spinal procedures, chemotherapy that brought fatigue and major weight loss, and a serious bout of pneumonia with atrial fibrillation, after which he chose to continue treatment anyway. The good news is that his stamina and walking have returned, and he's looking forward to swimming again. The unexpected subplot: while hospitalized, a newly announced PHP vulnerability forced him to secure his company's servers from his hospital bed late into the night. It's a candid story about resilience, and about how responsibility doesn't always wait until you're well.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily

- KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad

- Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad

Today's topics:

Python Drops Experimental JIT Compiler - CPython maintainers withdrew PEP 836, so Python's experimental JIT compiler will likely be removed from the main branch early in the Python 3.16 development cycle.

Ghostty 1.4.0 Terminal Release - Ghostty 1.4.0 adds full Kitty graphics protocol support, scrollback compression, a new Linux OpenGL renderer, Kitty clipboard support, faster performance, and several security fixes.

Tailscale SSH Privilege Escalation Flaws - Tailscale is patching privilege escalation vulnerabilities in Tailscale SSH and advises customers to review access policies and disable SSH forwarding they don't need.

Secret FISA Court Ruling Withheld - Senators Ron Wyden and Adam Schiff accuse the Trump administration of illegally withholding a FISA court ruling on Section 702 searches of Americans' communications ahead of the reauthorization debate.

DeepMind Unveils Genie 4 World Model - Google DeepMind's Genie 4 is a multimodal, real-time interactive world model built on Gemini, with better physics and natural-language steering, aimed at robotics training and entertainment.

Offline Survival AI Assistant - Survival LLM is a free, open-source, offline AI assistant for first aid and emergency questions that cites trusted sources such as FEMA, CDC and Wikipedia.

Version Control for US Law - A proposal to treat the US Code like source code, with full version history and effective dates, aims to make legal research and legal AI tools more accurate.

Inkjet Droplets Win Nikon Small World - German photographer Ole Bielfeldt won the 2026 Nikon Small World photomicrography competition with an image of inkjet printer droplets frozen in mid-flight.

-Ghostty 1.4.0 Brings Major Performance, Memory, and Security Upgrades

-Senators Accuse Trump Administration of Withholding Secret FISA Ruling

-Google DeepMind unveils Genie 4 for real-time world simulation

-Python’s experimental JIT is being pulled from CPython

-Apple unveils 14-inch MacBook Pro with M6 chip

-Survival LLM Brings Offline AI Help for Emergencies

-Apple Announces New M6 iMac with 24-Inch Retina Display

-Tailscale Investigates Security Flaws in Tailscale SSH

-Lexifina argues the U.S. Code should work like Git

-Ink Droplets Win 2026 Nikon Small World Contest

Episode Transcript

Python Drops Experimental JIT Compiler
Let's start with Python. The experimental JIT compiler that landed in CPython in 2024 is very likely being pulled from the main branch. The steering council had asked for a formal proposal, PEP 836, explaining how the JIT would be maintained, kept secure, and supported across platforms, and it set a six-month deadline. After community feedback and further concerns from the council, the maintainers withdrew the PEP, saying they simply don't have the bandwidth to carry it forward. The council had already said that without an accepted proposal the code would have to go, and the plan is to remove it early in the Python 3.16 cycle so any fallout surfaces quickly. The work may carry on outside the core repository, but for now, a headline effort to speed up Python from the inside has stalled. It's a reminder that in open source, maintainer time is usually the scarcest resource of all.

Ghostty 1.4.0 Terminal Release
Better news for developers who live in the terminal: Ghostty 1.4.0 is out, and it's a hefty release. The project says it now implements the entire published Kitty graphics specification, animation included. Scrollback is compressed transparently in the background, which saves enough memory that the default scrollback limit jumps fivefold on supported platforms. Performance improves across the board, especially on macOS, with faster launches, parsing, resizing, and search. Linux and FreeBSD users get a new OpenGL renderer that looks sharper under fractional scaling and tears less. There's also support for the Kitty clipboard protocol, so terminal apps can work with images and HTML on the clipboard. And worth flagging: the release includes several security fixes, some for malicious terminal output and graphics payloads. If you use Ghostty, this one is worth installing soon.

Tailscale SSH Privilege Escalation Flaws
On the security front, Tailscale says it is investigating vulnerabilities in Tailscale SSH, working with outside specialists to pin down the scope. Most of the issues involve privilege escalation, meaning a user who is allowed to SSH into a machine might be able to do more than their local permissions should allow, possibly up to elevated access. The company isn't telling everyone to switch the feature off, but it is treating the findings seriously. Most of the problems are already fixed, a patched client was targeted for the end of Tuesday, and a full advisory and technical write-up will follow. In the meantime, Tailscale suggests reviewing SSH access policies, turning off SSH forwarding where you don't need it, and tightening access to sensitive systems for now.

Secret FISA Court Ruling Withheld
Staying with security, but at the national level: Senators Ron Wyden and Adam Schiff are accusing the Trump administration of illegally withholding a secret ruling from the Foreign Intelligence Surveillance Court. According to WIRED, the ruling, issued about seven months ago, criticized how analysts use search filters to comb through Americans' communications collected under Section 702, and said those searches needed stronger legal safeguards. The court still approved another year of surveillance. The administration's own policy calls for such rulings to be reviewed for public release within 180 days, and that deadline passed in September. The timing matters because intelligence officials want Congress to renew Section 702 before the year ends, and the senators argue lawmakers can't make that call without seeing the ruling. They've asked for answers by November 6, and civil liberties groups are urging Congress to hold off on renewal until the ruling is public.

DeepMind Unveils Genie 4 World Model
Over to AI. Google DeepMind has announced Genie 4, its latest world model for real-time interactive simulations. The new part is that it generates visuals, ambient sound, music-like audio, and dialogue together, as one coherent world. Built on top of Gemini, it can be steered live with plain language, so you can change the weather, move the camera, or drop in new people and objects. DeepMind also says physics and human motion are more convincing. It starts as a research preview for trusted testers, with wider access planned later. The larger pitch is that models like this could become training grounds for robots and AI agents, not just a new kind of game.

Offline Survival AI Assistant
A very different take on AI comes from Survival LLM, a free, open-source assistant built to keep working when the internet doesn't. It runs entirely on your own computer and answers first-aid and survival questions from an offline library drawn from sources such as FEMA, the CDC, MedlinePlus, iFixit, and a large set of Wikipedia articles. Every answer comes with citations, which is a sensible design choice, because the project openly admits the model can still get things wrong. There are no accounts and no tracking, and it can also search files you add yourself.

Version Control for US Law
Sticking with accuracy, one essay argues that the U.S. Code should be handled like source code, with a full version history showing who changed a provision, when, and what it said on any given date. The author points out that language models often miss this timing and apply the wrong version of a statute. Bankruptcy rules are the example, where the right threshold depends on the filing date and on later changes that apply retroactively. The piece walks through one company's attempt to rebuild that history, and it's candid about the edge cases. It's an interesting case of git thinking applied to the law.

Inkjet Droplets Win Nikon Small World
To close, something lovely. A 25-year-old German experimental photographer, Ole Bielfeldt, won the 2026 Nikon Small World photomicrography competition, beating more than eighteen hundred entries from 69 countries. His winning image shows ink droplets landing on paper from an inkjet printer. He took the printer apart so the paper moved beneath a fixed printhead, then used flash in a dark room to freeze the droplets in mid-air. He got into microscopy during the COVID lockdown after finding an old microscope in his parents' attic, and taught himself by trial and error. No AI involved, just patience and a willingness to take things apart.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)

15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:

- Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily

- Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad

- SurveyMonkey, Using AI to surface insights faster and reduce manual analysis time - https://get.surveymonkey.com/tad

Today's topics:

Anthropic files S-1 for IPO - Anthropic filed a public S-1 to list Class A shares on Nasdaq under the ticker ANT. The filing details a Long-Term Benefit Trust, founder voting shares, and AI safety risk disclosures.

OpenAI claims AI-generated scientific breakthroughs - OpenAI says AI-generated results in math, computer science, and physics will be released and verified with independent academic working groups. The math result comes with a Lean proof, and OpenAI says it will take no academic credit.

Exponential Time Hypothesis claimed refuted - A paper with a Lean formalization claims the Exponential Time Hypothesis is false, using a subexponential, polynomial-space algorithm for counting k-CNF solutions. If it holds, many complexity theory lower bounds would need to be revisited.

CSS-Tricks returns to independence - DigitalOcean sold CSS-Tricks to Geoff Graham, making the front-end publication independent again. Guest writing, the newsletter, and more beginner-friendly content are planned, funded by ads and memberships.

Oldest app still running on Windows 11 - A retro computing investigation found that CDPLAYER.EXE from November 1992 is the oldest GUI app that runs natively on Windows 11. A 1993 Win32s FreeCell also still works, a strong example of Windows backward compatibility.

Solar power overtakes nuclear globally - Solar power is overtaking nuclear and is expected to supply about 10% of global electricity in 2026. Cheap panels from China, battery storage, and adoption in India, Pakistan, and Australia are driving growth, while grids need to keep up.

Designing websites that print well - A web developer argues for print-friendly websites built with CSS @media print rules. The goal is pages that are readable in black and white and easy to annotate offline.

Pascal's 1662 Paris transit network - Blaise Pascal's 1662 horse-carriage service in Paris introduced fixed-route public transit on land. The article argues that shared transit networks still beat on-demand travel and the hype around driverless cars.

Microsoft's Type Server Protocol for Python - Microsoft documented the Type Server Protocol, a JSON-RPC spec that lets Python tools query a type server for type and import data. It builds on Language Server Protocol conventions and supports parallel read-only connections.

-Anthropic Files S-1 for IPO Under Ticker ANT

-OpenAI Launches Academic Collaborations for AI-Generated Discoveries

-Sebastian Raschka Launches a Full Course on Building Reasoning Models

-OpenAI Paper Claims ETH Is False

-CSS-Tricks Returns to Independent Ownership

-Oldest Native GUI App That Runs on Windows 11

-Solar Power Overtakes Nuclear as Costs Plunge and Adoption Accelerates

-Why Websites Should Be Printable

-Why Blaise Pascal Invented Modern Public Transit

-Type Server Protocol Overview

Episode Transcript

Anthropic files S-1 for IPO
First, an update on Anthropic. The company has now filed a public S-1 registration statement, the formal start of its initial public offering. It plans to list on Nasdaq under the ticker ANT. The prospectus spells out an unusual governance setup. A Long-Term Benefit Trust holds special Class T shares that carry no economic rights, yet they let the trust appoint a majority of the board. Founders keep extra voting power through a separate Class F structure. The risk section is also notable. Alongside the usual business warnings, Anthropic says advanced AI could become difficult for the company itself to evaluate and control. That is an unusual thing to tell prospective shareholders, and it shows how the company is trying to balance fast growth with public-benefit oversight.

OpenAI claims AI-generated scientific breakthroughs
Now, the story from the top of the show. OpenAI says it is changing how it releases major research produced by its AI systems. It reports three results. The first is a mathematics proof related to the local smoothing and Kakeya conjectures. The second is a refutation of the Exponential Time Hypothesis in computer science. The third is an existence proof for certain quantum field theories connected to the Standard Model. OpenAI calls these outputs raw research data that still need scientific interpretation. The math result went out publicly with a plain-language writeup and a machine-checked proof in Lean. The computer science and physics results are being handled by independent academic working groups. OpenAI says it will take no academic credit, will credit the collaborators, and will fund efforts to help researchers build on the work.

A paper making the Exponential Time Hypothesis claim has also surfaced, together with its Lean formalization. It describes a deterministic algorithm that counts the solutions to a classic family of logic problems in less than exponential time, using only a modest amount of memory. This hypothesis is the baseline that many hardness results and algorithmic lower bounds rest on. If the claim survives scrutiny, a large body of theoretical work would need revisiting. If it doesn't, we will learn a lot about how to vet machine-generated proofs. Either way, verification is now the central question.

Exponential Time Hypothesis claimed refuted
On to web publishing. DigitalOcean has sold CSS-Tricks back into independent hands. Geoff Graham, a longtime contributor, is taking it over as his full-time job with a small team, four years after DigitalOcean bought the site from founder Chris Coyier. New articles will keep coming, though probably at a slower pace. The guest writing program and the newsletter are returning, and Graham wants the site to be friendlier to beginners. Funding will come from ads, memberships, a shop, and eventually courses. It is a small but welcome sign that community-run publishing still has a path forward.

CSS-Tricks returns to independence
Staying with front-end work, one developer makes the case that websites should print well. When they printed their own site, the result was oversized, ugly, and hard to read in black and white. A set of print-specific CSS rules fixed it: hiding page clutter, adjusting type, and making links and code blocks readable on paper. The author admits it took more effort than expected, but calls it a small kindness to readers who still like to mark up pages with a pen.

Oldest app still running on Windows 11
Here's a nice bit of software archaeology. One researcher went looking for the oldest graphical app that runs on Windows 11 with no emulation at all. A FreeCell build from February 1993 still launches and plays fine. Digging through early Windows NT builds, the author found that many older programs crash because they used draft executable formats or outdated conventions. The oldest survivor is the CD Player app from November 1992. It still works because a system library it depends on still ships with Windows. With emulators, you can even run Reversi from Windows 1.01, though the author fairly calls that cheating. It is a good reminder of how far Microsoft's compatibility promise really reaches.

Solar power overtakes nuclear globally
Energy next. Solar power is now overtaking nuclear worldwide and is expected to supply about a tenth of global electricity in 2026. A decade ago it was a marginal source, and forecasters have underestimated it again and again. Huge manufacturing scale, mostly in China, drove panel prices down. Today solar is the cheapest new power in many markets, and adoption is surging in places like India, Pakistan, and Australia. Batteries are extending solar into the evening, with California a leading example. The bottleneck now is less the panels and more whether grids, storage, and institutions can keep pace.

Designing websites that print well
Next, a history lesson with a modern edge. In 1662, Blaise Pascal launched a horse-carriage service in Paris built around fixed routes and predictable frequencies. The article argues this was essentially the first land-based public transit network. It used radial and cross-town routes that planners still rely on today. It was cheap, popular, and brought different social classes into the same carriages. The author's larger point is that shared, fixed-route transit solves a basic problem of geometry in crowded cities, something hype about driverless cars tends to forget.

Pascal's 1662 Paris transit network
Finally, for Python tooling fans, Microsoft has documented the Type Server Protocol. It is a standard way for editors and tools to ask a dedicated type server about types and imports. It builds on familiar Language Server Protocol conventions and allows extra read-only connections so requests can run in parallel. It's still pre-1.0, so breaking changes are possible, but it hints at more modular type checking across Python editors.

Subscribe to edition specific feeds:

- Space news

* Apple Podcast English

* Spotify English

* RSS English Spanish French

- Top news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- Tech news

* Apple Podcast English Spanish French

* Spotify English Spanish Spanish

* RSS English Spanish French

- Hacker news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

- AI news

* Apple Podcast English Spanish French

* Spotify English Spanish French

* RSS English Spanish French

Visit our website at https://theautomateddaily.com/

Send feedback to [email protected]

Youtube

LinkedIn

X (Twitter)