Daily Prompt with Archer and Iris - Daily AI News

Daily Prompt with Archer & Iris

Ep. 104 — Grok 4.5 goes after Anthropic's coding turf; OpenAI's voice learns to shut up

Thursday, July 9, 2026

Today's stories:

LEADS

- Grok 4.5 lands "Opus-class": SpaceXAI's first model since its IPO and Cursor buy is a coding-focused model Elon Musk calls "Opus-class, but faster and lower cost" — $2/$6 per million tokens vs Anthropic Opus at $5/$25.

https://techcrunch.com/2026/07/08/spacexai-releases-grok-4-5-which-elon-describes-as-an-opus-class-model/

- OpenAI's GPT-Live voice models: New full-duplex GPT-Live-1 and GPT-Live-1 mini speak and listen at once; mini replaces Advanced Voice as the ChatGPT default. Built to stay quiet — "better at shutting up" — and targets 30-40 minute conversations.

https://techcrunch.com/2026/07/08/openai-releases-new-voice-models-for-more-natural-live-conversations/

- Lovable reportedly raising at $13.2B: The Swedish vibe-coding startup would double its December valuation on a $500M revenue run rate. Mark Cuban argues the labs won't swallow the app layer because the model was never the hard part.

https://techcrunch.com/2026/07/08/lovable-reportedly-in-talks-to-double-its-valuation-to-13-2b/

QUICK HITS

- Jensen Huang says his Nvidia engineers would rather build agents than write code — and love the shift.

https://www.businessinsider.com/jensen-huang-my-engineers-prefer-building-agents-to-writing-code-2026-7

- Google's SynthID watermark helped debunk a fake image of Senator McConnell in a hospital bed — but Anthropic doesn't participate in the program.

https://techcrunch.com/2026/07/08/googles-deepfake-detector-system-used-to-debunk-mcconnell-hoax-pic/

- Meta paused its "Model Capability Initiative" — which trained AI on employee keystrokes — after data ended up where it shouldn't.

https://www.businessinsider.com/meta-andrew-bosworth-why-ai-training-employee-keystrokes-paused-2026-7

- Perplexity, valued at $20B, is quietly building a coding tool called "Teammate" to take on Cursor and Claude Code.

https://www.businessinsider.com/perplexity-building-ai-coding-tool-take-on-cursor-and-openai-2026-7

Daily Prompt with Archer & Iris

Ep. 103 — Government clears GPT-5.6; Meta's Muse pulls strangers into AI photos

Wednesday, July 8, 2026

Today's stories:

- The Trump administration lifts restrictions on OpenAI's GPT-5.6. The Commerce Department cleared a broad public launch after CAISI testing; the flagship "Sol" and lower tiers "Terra" and "Luna" go public Thursday. Governments are now negotiating frontier-model access case-by-case, one release at a time.

https://www.axios.com/2026/07/08/openai-gpt-trump-ban-lifted

- Meta launches Muse Image, the first model from its Superintelligence Labs, now powering Instagram, WhatsApp, and the Meta AI app. Backlash centers on a feature that pulls public Instagram profiles into AI images without explicit consent — opt-out by default, no notification.

https://www.theverge.com/tech/962485/meta-muse-image-ai-model-instagram

- Anthropic brings Claude Cowork to mobile and web (Max subscribers first), enabling async office tasks across devices — and to government users. Anthropic's own usage data across 1.2M sessions and 600,000+ orgs: business process automation 33.4%, content 16.4%, software development just 8.7%.

https://techcrunch.com/2026/07/07/the-coding-agent-wars-are-spilling-into-the-rest-of-the-office-claude-cowork/

Quick hits:

- SambaNova raises $1B at an $11B valuation (Series F first close, led by General Atlantic), months after a rumored ~$1.6B Intel acquisition stalled.

https://techcrunch.com/2026/07/08/sambanova-draws-1b-at-11b-valuation-in-series-f-first-close/

- French startup ZML releases a free inference server, LLMD, that runs LLMs across Nvidia, AMD, Google TPU, Apple Metal, and Intel Arc chips. Endorsed by Yann LeCun.

https://techcrunch.com/2026/07/08/hot-french-startup-zml-releases-free-product-to-speed-inference-across-lots-of-ai-chips/

- Google expands Managed Agents in the Gemini API: background tasks, direct remote MCP connections, custom function calling, and credential refresh.

https://blog.google/innovation-and-ai/technology/developers-tools/expanding-managed-agents-gemini-api/

- Discord admits an AI moderation bug wrongfully banned users over harmless images since May, with ~200 more banned over the weekend before the fix.

https://techcrunch.com/2026/07/07/discord-admits-ai-moderation-bug-wrongfully-banned-users-over-harmless-images/

Ep. 102 — Microsoft cuts 4,800 and the first AI-run ransomware still needed a human

Daily Prompt with Archer & Iris

Tuesday, July 7, 2026

Today's theme is the human still standing in the loop. Machines can now carry more of the job than ever — the layoffs, the heist, the migration, the moderation — yet every story lands on a person doing the one part that's left: deciding. Microsoft cuts jobs while insisting AI isn't why, a security firm catches ransomware that ran itself but still needed a human to aim it, and Vercel's CEO argues the model should be a swappable part, not a marriage. Plus five quick hits.

LEADS

- Microsoft laid off 4,800 employees (about 2.1% of its workforce), including 1,600 from Xbox in what its CEO called "the most significant restructure in Xbox history," citing margins "3 to 10x lower" than comparable platforms. HR chief Amy Coleman said the roles "are not being replaced by AI" — while noting AI "is changing how work gets done." It follows ~15,000 cuts in 2025, spring buyouts to ~5,500, and a $2.5B AI deployment unit.

https://techcrunch.com/2026/07/06/microsoft-lays-off-nearly-5000-employees-across-xbox-commercial-sales/

- Security firm Sysdig documented what it calls the first ransomware run largely by an AI: it exploited a flaw in an open-source LLM app builder, stole credentials, moved through the network, encrypted 1,300+ records, and wrote its own ransom note with a Bitcoin address — even fixing a failed login in 31 seconds. But a human still provisioned the infrastructure and chose the victim. The warning: attacks like this are now bounded by budget, not effort.

https://techcrunch.com/2026/07/06/the-first-ai-run-ransomware-attack-still-needed-a-human/

- Vercel CEO Guillermo Rauch argues the era of committing to a single AI lab is over: models and agents should be decoupled so teams can swap models on price/performance. He points to quiet growth in Google's model and open models (DeepSeek, GLM) showing up in production — a builder-strategy case for treating the model as a component, not a commitment.

https://techcrunch.com/2026/07/06/vercel-ceo-guillermo-rauch-on-the-fight-to-split-off-models-from-agents/

QUICK HITS

- Anthropic published a field guide to Claude Fable 5, its most capable model, centered on surfacing your own "unknowns" — arguing the quality of agentic work is now bottlenecked by how well you can clarify what you don't know you know.

https://claude.com/blog/a-field-guide-to-claude-fable-finding-your-unknowns

- Apple's iOS 27 beta 3 turns on new Siri controls for "Pace" and "Expressivity," letting users tune how fast and how emotional the LLM-rebuilt assistant sounds, with live previews.

https://techcrunch.com/2026/07/06/you-can-now-customize-siris-pace-and-expressivity-in-the-latest-ios-27-beta/

- Reddit says its LLM-based tools now block 23 million spam views per day — fighting the exact category of automated content that cheap language models made easy.

https://techcrunch.com/2026/07/06/reddit-is-using-llms-to-solve-a-problem-llms-largely-created/

- MIT Technology Review runs the math on OpenAI's floated 5% government stake: at an $852B valuation split across ~133 million households, it comes to about $320 each.

https://www.technologyreview.com/2026/07/06/1140176/your-familys-300-stake-in-openai/

- Forterra's autonomous ATVs are running missions in Ukraine — 100+ vehicles, 2,500+ miles, 52 casualty evacuations — but crews mostly teleoperate because, per a soldier, the autonomy "doesn't know how to" react to live threats yet.

https://techcrunch.com/2026/07/07/the-first-american-autonomous-ground-vehicles-are-fighting-in-ukraine/

Ep. 101 — Amazon closes Mechanical Turk; Anthropic starts making its own drugs

Daily Prompt with Archer & Iris

Monday, July 6, 2026

The weekend's theme was the machines reaching into the human parts of AI — the labeling, the discovery, the taste. Amazon winds down the crowdwork marketplace that trained a decade of machine learning (right as a third-to-half of its workers were secretly using LLMs to do the "human" tasks), Anthropic stands up its own in-house drug lab, and Google DeepMind moves in with film studio A24. Plus four quick hits.

LEADS

- Amazon will stop accepting new customers for Mechanical Turk as of July 30, 2026 — existing accounts stay open, but no new features are planned. Launched in 2005 and repositioned around 2018 as a data-labeling tool, MTurk was the invisible human layer under modern AI. The kicker: a 2023 analysis found between a third and nearly half of its workers were already using large language models to complete their tasks — the humans hired to teach the machines were quietly asking the machines.

https://techcrunch.com/2026/07/05/amazon-will-stop-accepting-new-customers-for-mechanical-turk/

- Anthropic is starting its own in-house drug discovery program — not just selling Claude Science to pharma, but developing treatments itself, focused on neglected and rare single-gene diseases big pharma won't touch. Led by life-sciences head Eric Kauderer-Abrams, who says the company needs "tight feedback loops" and hands-on experience to build better tools. Backed by a ~$400M acquisition of Coefficient Bio and Novartis CEO Vas Narasimhan joining the board. All preclinical; no disclosed targets or commercialization plans.

https://www.cnbc.com/2026/06/30/anthropic-launches-ai-drug-discovery-program-claude-science.html

- Google DeepMind and A24 announced a "first-of-its-kind" research partnership, with Google investing in the studio and filmmakers working alongside researchers to shape the technology. Notably, the announcement names no product, no video model, and no release date — reading less like a tool launch and more like DeepMind buying creative taste and A24 buying an early seat at the lab while rival studios litigate.

https://deepmind.google/blog/google-deepmind-and-a24-announce-first-of-its-kind-research-partnership/

QUICK HITS

- Alibaba is banning employees from using Anthropic's Claude Code as of July 10, flagging it as high-risk over a March feature that could identify Chinese users, and directing staff to its in-house Qoder tool. Anthropic's Thariq Shihipar called the feature an anti-reseller/anti-distillation experiment it had meant to remove; Anthropic already bars Chinese firms.

https://techcrunch.com/2026/07/04/alibaba-reportedly-bans-employees-from-using-claude-code/

- OpenAI hired Paul Meade, Apple's Vision Pro hardware lead of seven years, to work on its upcoming family of AI devices alongside Jony Ive's team (the io startup OpenAI acquired for $6.5B last year).

https://finance.yahoo.com/technology/ai/articles/apple-vision-pro-smart-glasses-193101024.html

- Midjourney is demanding that Disney, Universal, and Warner Bros. disclose their own internal AI usage — every prompt and model — arguing that if the studios suing it build image models on unlicensed content, that establishes industry custom supporting a fair-use defense.

https://techcrunch.com/2026/07/04/midjourney-wants-hollywood-studios-to-reveal-the-details-of-their-ai-usage/

- Google's July 4th commercial imagined the Founding Fathers drafting the Declaration of Independence in Google Workspace with Gemini — drawing backlash for being tone-deaf, with critics noting the ad about human creativity appeared partly AI-generated.

https://techcrunch.com/2026/07/04/new-google-commercial-imagines-a-declaration-of-independence-written-with-help-from-ai/

Deep Dive: Scale AI (Part 2 — Did Meta Buy a Genius or an Invoice?)

Daily Prompt with Archer & Iris

Sunday, July 5, 2026

Part two of the weekend deep dive on Scale AI. Yesterday: the cast, the empire, and the $14.3B Meta deal that broke the company's neutrality. Today: the live stakes. What Scale is becoming without its old identity, whether the acqui-hire of Alexandr Wang is paying off inside Meta, and what builders should take from a market where the industry's neutral supplier picked a side.

IN THIS EPISODE

- Scale's reinvention: New CEO Jason Droege is flipping the revenue mix from ~70% data labeling toward enterprise apps (already ~$200M/yr, projected to overtake labeling within 18 months). Scale did just shy of $1B in 2025 (up from $870M in 2024) and projects $1B+ in 2026, possibly near $2B. Meta is the anchor — committed to at least $450M/year for five years, nearly half of Scale's revenue. New clients include Mayo Clinic, BP, Allianz, and Cisco. The biggest new bet is defense: a $500M Pentagon mission-planning program and a missile-defense role alongside Palantir and Anduril. Headcount ~1,300, hiring 500 more this year.

- The reframe: Droege didn't win back the customers who fled — he pivoted to the two buyers who structurally can't leave. Meta is locked into a five-year contract; the Pentagon doesn't shop your cap table. Reinvention from the outside, captive customers from the inside — and the neutral supplier has effectively become a defense contractor.

- Wang inside Meta: Alexandr Wang, 28, is Chief AI Officer of Meta's superintelligence effort, running the team behind its biggest models. Meta has spent $100B+ on AI infrastructure since early 2025, including $30B+ in a single quarter. The lab shipped its first major model under Wang in April 2026, powering the Meta AI assistant with a parallel-agent reasoning mode. But the environment is turbulent: Meta's star chief scientist departed in November 2025, and the lab cut ~600 jobs in fall 2025. One framing: Zuckerberg asked Wang to "build the engine while figuring out what car to put it in." The $14.3B bet is still uncalled — genius or expensive employee.

- Builder takeaways: The human-feedback and evaluation layer under every model is a strategic dependency, not a commodity. The single-supplier concentration broke when Scale picked a side; demand scattered to rivals (Handshake, Labelbox, Surge, Mercor) and some labs are bringing evals in-house. The real emerging moat is the evaluation layer — the judgment around the model, not the model itself. The lesson: the most valuable thing in your supply chain is often the least technical thing in it, and trust doesn't transfer when a company changes hands.

NEXT WEEKEND

- Deep Dive: CoreWeave — the GPU landlord under the whole AI boom.

SOURCES

- Scale AI one year after the Meta deal: Droege's pivot to apps and defense (Forbes)

https://www.forbes.com/sites/richardnieva/2026/05/14/scale-meta-deal/

- A year after Meta tapped Wang to build a new AI model, Zuckerberg has to sell it (CNBC)

https://www.cnbc.com/2026/06/14/meta-hired-alexandr-wang-to-build-ai-its-zuckerbergs-job-to-sell-it.html

- Meta debuts Muse Spark, its first major model since the $14B Scale deal (CNBC)

https://www.cnbc.com/2026/04/08/meta-debuts-first-major-ai-model-since-14-billion-deal-to-bring-in-alexandr-wang.html

- Meta Superintelligence Labs — background (Wikipedia)

https://en.wikipedia.org/wiki/Meta_Superintelligence_Labs

- The inside story of Alexandr Wang's rise and the $14B Meta deal (Fortune)

https://fortune.com/2025/06/22/inside-rise-scale-alexandr-wang-meta-zuckerberg-14-billion-deal-acquihire-ai-supremacy-race/

Ep. 99 — Deep Dive: Scale AI (Part 1 — The Company Every Model Was Built On)

Daily Prompt with Archer & Iris

Saturday, July 4, 2026

Almost every AI you've ever used was quietly built on top of one company most people can't name: Scale AI. Part 1 of a two-part weekend deep dive on the data-labeling empire, its 28-year-old founder, and the Meta deal that broke the one thing the company was actually selling — its neutrality. Tomorrow, Part 2: the live stakes.

IN THIS EPISODE

- The cast: Alexandr Wang — Los Alamos physicist parents, national math competitions, one year at MIT before dropping out, co-founding Scale AI with Lucy Guo in 2016 through Y Combinator (after a doctors' concierge idea went nowhere), and becoming the youngest self-made billionaire at 24. Guo, pushed out in 2018, still collected a reported $750 million from the Meta deal.

- The business: the human layer under modern AI — roughly 60,000 contractors labeling data, ranking answers, and doing RLHF (reinforcement learning from human feedback), which Scale pioneered with OpenAI in 2019. Wang's pitch was neutrality: "nearly every major model is built on our data foundry." (The U.S. Labor Department opened a worker-pay/classification probe in 2024; it closed in 2025.)

- The deal & the collapse: Meta's June 2025 investment — $14.3 billion for a 49% non-voting stake, valuing Scale near $29 billion — and the 1.5x share-conversion "poison pill" that reveals it as an acqui-hire of Wang, who left to run Meta's superintelligence lab. The moment the deal landed, Scale's neutrality evaporated: Google (its largest customer, ~$200M/yr), OpenAI, Microsoft, and xAI pulled back, while rivals like Handshake ("demand tripled overnight") and Labelbox feasted.

- The bridge to Part 2: new CEO Jason Droege pivots Scale toward apps and defense — a ~$500M Pentagon program (Project Thunderforge) and a missile-defense role alongside Palantir and Anduril.

SOURCES

- The inside story of Alexandr Wang's rise and the $14B Meta deal (Fortune)

https://fortune.com/2025/06/22/inside-rise-scale-alexandr-wang-meta-zuckerberg-14-billion-deal-acquihire-ai-supremacy-race/

- Scale AI's Wang confirms departure for Meta as part of $14.3B deal (CNBC)

https://www.cnbc.com/2025/06/12/scale-ai-founder-wang-announces-exit-for-meta-part-of-14-billion-deal.html

- Google, Scale AI's largest customer, plans split after Meta deal (CNBC)

https://www.cnbc.com/2025/06/14/google-scale-ais-largest-customer-plans-split-after-meta-deal.html

- Inside Scale AI's business after Meta's $14 billion deal (Forbes)

https://www.forbes.com/sites/richardnieva/2026/05/14/scale-meta-deal/

- A year after Meta tapped Wang, Zuckerberg has to sell it (CNBC)

https://www.cnbc.com/2026/06/14/meta-hired-alexandr-wang-to-build-ai-its-zuckerbergs-job-to-sell-it.html

NEXT

Tomorrow — Part 2: Scale after Wang, the live stakes, and the real question: did Meta buy intelligence, or a very expensive employee?

Ep. 98 — OpenAI offers the U.S. a 5% stake, and Meta's agents run late

Daily Prompt with Archer & Iris

Friday, July 3, 2026

Today's episode is about the bill for building AI — who pays, who owns, and who's admitting it's slower than promised. OpenAI floats handing the government a stake, Anthropic goes shopping for its own silicon, and Zuckerberg tells staff the agent revolution is running behind schedule.

LEADS

- OpenAI proposed donating 5% of its equity to a U.S. sovereign wealth fund, with other AI companies contributing similarly — a voluntary offer that reads as a counter to Senator Sanders' proposed one-time 50% tax on AI stock. President Trump confirmed the talks; it's preliminary and would need Congress.

https://techcrunch.com/2026/07/02/openai-proposed-donating-5-of-its-equity-to-a-us-sovereign-wealth-fund/

- Anthropic is in early talks with Samsung to build a custom AI chip — not to leave Nvidia, but to diversify a stack that already spans Google, Amazon, and Nvidia. Follows OpenAI's own custom inference chip with Broadcom. The signal: the binding constraint has moved from the model to the metal.

https://techcrunch.com/2026/07/02/anthropic-is-discussing-a-new-custom-chip-with-samsung/

- Mark Zuckerberg told Meta staff at a town hall that AI agents haven't "accelerated in the way" the company expected — after cutting ~8,000 jobs (10%), reassigning 7,000 into AI teams, and committing as much as $145B to AI infrastructure in 2026. His fix: three to six months.

https://techcrunch.com/2026/07/02/mark-zuckerberg-tells-staff-that-ai-agents-havent-progressed-as-quickly-as-hed-hoped/

QUICK HITS

- Microsoft launched Frontier, a $2.5B AI-deployment company with 6,000 engineers — the third major "make the AI deal actually work" org in a week (after AWS's $1B on June 30).

https://techcrunch.com/2026/07/02/microsoft-launches-its-own-ai-deployment-company-with-2-5-billion-commitment/

- Anthropic shipped admin tools for Claude spend: per-user and per-team cost visibility, spend alerts at 75% and 90% of budget, and an analytics API — a sign agentic usage now needs a finance dashboard.

https://claude.com/blog/giving-admins-more-visibility-and-control-over-claude-usage-and-spend

- Meta quietly launched Pocket, a "vibe-coded" app that turns a text prompt into a tiny game — a rebrand of the acquired Gizmo (635,000 lifetime installs) — the same week it admitted agents are slow.

https://techcrunch.com/2026/07/02/meta-quietly-launches-vibe-coded-gaming-app-pocket/

- Factory released Droid Shield 2.0, a secret-detection system that stops autonomous coding agents from committing passwords and API keys — with open-source LoRA adapters on Hugging Face.

https://factory.ai/news/droid-shield-2-0

Daily Prompt with Archer & Iris

Nobody shipped a model today — everyone grabbed a layer

Thursday, July 2, 2026

Nobody released a new frontier model, and that was the story. Archer and Iris cover a day where the big moves were about position, not intelligence: Anthropic building scientists a workbench, Cloudflare drawing a toll line across the web, and Google planting an always-on agent on your desktop — plus four quick hits.

Leads

- Anthropic launches Claude Science — a research workbench (not a new model) that unifies 60+ databases, a project-manager assistant, sub-assistants, and a dedicated citation/math fact-checker. Early users: Novo Nordisk, the Allen Institute. On-prem option; 50 funded projects (up to $30K credits, apply by July 15).

https://techcrunch.com/2026/06/30/anthropics-claude-science-bets-on-workflow-not-a-new-model-to-win-over-scientists/

- Cloudflare pushes AI companies to pay for publishers' content — mixed-use crawlers blocked by default from Sept 15 unless search and training/agent crawlers are separated, plus a new pay-per-use model. CEO Matthew Prince: "the majority of traffic on the Internet is non-human."

https://techcrunch.com/2026/07/01/cloudflares-new-policy-pushes-ai-companies-to-pay-for-publishers-content/

- Gemini Spark, Google's 24/7 agentic assistant, arrives on Mac — continuous real-time tracking and autonomous errands (files, invoices-to-budgets, reservations, groceries). Beta, AI Ultra only, US only.

https://techcrunch.com/2026/07/01/gemini-spark-googles-agentic-assistant-is-now-available-on-mac/

Quick hits

- OpenAI releases GeneBench-Pro, a benchmark for real-world genomics and biology — the same week Anthropic ships a science workbench.

https://openai.com/index/introducing-genebench-pro

- SpaceX has a prototype AI device, handset-shaped and wired to xAI's OS; Musk calls the reports "utterly false."

https://techcrunch.com/2026/07/01/spacex-has-an-ai-device-prototype-and-it-sure-sounds-phone-ish/

- Meta plans "Meta Compute," a cloud business selling raw compute and its own models, after committing $182.9B to AI infrastructure.

https://techcrunch.com/2026/07/01/meta-like-spacex-looks-to-turn-excess-ai-compute-into-cash/

- Venice AI hits unicorn status with a $65M Series A at a $1B valuation — profitable, privacy-first, stores no user data.

https://techcrunch.com/2026/07/01/venice-ai-becomes-a-unicorn-with-65m-series-a-as-its-privacy-first-ai-platform-takes-off/

Daily Prompt with Archer & Iris

Claude Sonnet 5, three-cent images, and a $5B inference chip

Wednesday, July 1, 2026

The whole stack got cheaper on the same day. Archer and Iris cover a near-flagship agent model priced to leave running, image generation that rounds down to a rumor, and a chip startup betting that inference is the bill that never stops — plus five quick hits.

Leads

- Anthropic launches Claude Sonnet 5 — near-Opus agentic ability at midsize price ($2/$10 per million tokens through Aug 31), now the default for free and Pro tiers.

https://techcrunch.com/2026/06/30/anthropic-launches-claude-sonnet-5-as-a-cheaper-way-to-run-agents/

- Google ships Nano Banana 2 Lite (four-second images, $0.034 per 1,000) and Gemini Omni Flash video ($0.10/sec).

https://techcrunch.com/2026/06/30/google-introduces-a-faster-cheaper-image-generator-with-nano-banana-2-lite/

- Nvidia challenger Etched hits a $5B valuation and $1B in booked orders for a transformer-only inference chip.

https://techcrunch.com/2026/06/30/nvidia-competitor-etched-hits-5b-valuation-1b-in-sales-for-ai-chip/

Quick hits

- Trump administration drops export restrictions on Anthropic's Mythos and Fable models; access returns July 1.

https://techcrunch.com/2026/06/30/trump-drops-restrictions-on-anthropics-mythos-and-fable-models/

- Meta's Brain2Qwerty decodes typed sentences from non-invasive MEG brain scans — 61% average word accuracy (78% best), up from 8%.

https://ai.meta.com/blog/brain2qwerty-brain-ai-human-communication

- Amazon launches a $1B forward-deployed-engineer org, following OpenAI ($4B) and Anthropic ($1.5B).

https://techcrunch.com/2026/06/30/amazon-launches-new-1-billion-fde-org-following-openai-and-anthropic/

- X launches a hosted, read-only MCP server for AI tools like Claude, Cursor, and Grok Build.

https://techcrunch.com/2026/06/30/x-now-offers-an-mcp-server-to-make-its-platform-easier-for-ai-tools-to-use/

- OpenClaw, the viral free open-source agent, ships iOS and Android apps.

https://techcrunch.com/2026/06/30/openclaw-is-finally-available-on-android-and-ios/

Gemini 3.5 goes agentic, the $900B memory crunch, and half-price Claude

Daily Prompt with Archer & Iris — Tuesday, June 30, 2026 (Daily Briefing)

Google shipped a new frontier model that doesn't just answer — it acts. But the louder story is everyone else fighting over the rest of the stack: the chips underneath it, the governments on top of it, and the models in the middle. The thread today is ownership.

LEAD STORIES

- Google ships Gemini 3.5, pitched as "intelligence with action." You hand it a goal; it plans the steps and spins up parallel subagents under a supervisor — maintaining a codebase, prepping financial documents, sorting assets. Google says it beats its own last flagship on coding and agentic tasks and outputs about 4x faster than other frontier models. Free in the Gemini app and Search, and on the developer API day one. The catch builders should sit with: when the model takes the action, a mistake is no longer a wrong sentence — it's a wrong action, already taken.

https://deepmind.google/blog/gemini-3-5-frontier-intelligence-with-action/

- South Korea commits $900B+ to ease "RAMageddon" — an AI-driven global memory shortage. The plan includes about $518B for four new memory fabs (plus ~$52B for an HBM packaging hub and ~$356B for AI data centers through 2035). President Lee called semiconductors, physical AI, and data centers the "triple axis," noting existing plants have "already reached their limits." The market already moved: memory maker Micron's stock jumped 236% in a month, revenue quadrupled to ~$41.45B in a quarter, and its market cap briefly topped Meta and Tesla — earning the "next Nvidia" label. The shortage runs into 2027. The bottleneck moved from the chip that thinks to the memory that feeds it — and it's spilling into laptop and phone prices.

https://techcrunch.com/2026/06/29/south-korean-tech-giants-commit-over-550b-to-ease-ramageddon/

https://techcrunch.com/2026/06/28/why-wall-street-thinks-us-memory-maker-micron-is-the-next-nvidia/

- Anthropic and Gov. Newsom give California state agencies and local governments Claude at half price. Newsom: AI "should not replace the human work of government; it should help our workers move faster... and deliver better results for Californians." The tell: Anthropic is the vendor the federal government has handled more cautiously on supply-chain risk, yet California's CIO said that designation "just didn't come up" — while the Pentagon's marquee AI deal went to OpenAI. Same company, two governments, opposite risk readings — and states are the next anchor customer the labs are racing to sign.

https://techcrunch.com/2026/06/29/anthropic-and-gov-newsom-forge-deal-allowing-california-government-to-use-claude-at-half-price/

QUICK HITS

- Vibe-coding platform Base44 (acquired by Wix for $80M last year) launched its own model, Base1, trained on tens of millions of user sessions — aiming to be cheaper and faster than renting frontier models. Base44 is at $100M run-rate revenue; rival Lovable is at $500M.

https://techcrunch.com/2026/06/29/vibe-coding-platform-base44-launches-own-model-as-ai-startups-seek-defensibility/

- Microsoft launched MAI-Code-1-Flash, a lightweight agentic coding model native to VS Code and Copilot — another first-party model that does OpenAI's job, hedging Microsoft's biggest partnership.

https://microsoft.ai/models/mai-code-1-flash/

- Crypto exchange OKX opened OKX AI, a marketplace where AI agents hire and pay each other in stablecoins around the clock, with on-chain reputations. It launched to developers after a 50-provider beta; the CEO is betting on million-dollar one-person companies.

https://techcrunch.com/2026/06/30/crypto-exchange-okx-wants-ai-agents-to-hire-and-pay-each-other/

- Arena, the AI leaderboard everyone uses, is now a $100M run-rate business — eight months after it started charging, on 10M+ human votes. The CEO's caveat: the revenue "is not recurring," and most people still think it's a free open-source project.

https://techcrunch.com/2026/06/29/arena-the-ai-leaderboard-everyone-uses-is-now-a-100m-business/