
Sign up to save your podcasts
Or


Watch the video version: https://www.youtube.com/watch?v=G4cLP0LU4I4
AI News Today: Gemini 4 Hydrogen, ChatGPT Wallet and a Free AI Brain
AI can now shop for you, read huge files for a flat price, and be downloaded for free. Today on The Daily Neural, Alex and Sam break down the three stories that matter.
Story 1: Google launches Gemini 4 Hydrogen, a fast new model with flat pricing across a huge memory for long documents.
Story 2: OpenAI rolls out ChatGPT Wallet, letting its AI helpers buy online with your fingerprint approval and a stand in card number that hides your real card.
Story 3: Thinking Machines Lab releases the downloadable brain of Inkling 2 Small for free, after a week of outside safety checks.
Would you let an AI helper shop for you? Tell us in the comments and subscribe for daily AI news in plain English.
#ainews #gemini #openai #chatgpt #artificialintelligence #technews #google #aiagents #opensource #thedailyneural
Watch the video version: https://www.youtube.com/watch?v=WRNp0ntMpxI
AI News Today: Google Gemini 4 Hydrogen, Claude Agents Go Private, Lawsuit to Freeze AI
Google just put a faster and cheaper AI model inside Search for everyone. Anthropic is letting companies run AI agents on their own servers. And four nonprofits are suing the biggest AI labs to freeze new models in court.
Alex and Sam break down all three stories in plain English, with every technical term explained the first time it shows up.
Today on The Daily Neural:
1. Google launches Gemini 4 Hydrogen. Built for everyday coding and for AI agents that do multi step tasks on their own, it costs $1.50 per million words in and $7.50 per million words out, with the same price across its full memory. It scored 51 on a widely watched test at its default setting and 54 turned up, just ahead of the larger Argon model, while finishing tasks faster. It is rolling out in AI Studio, the Gemini app, business tools, and Google Search.
2. Anthropic opens self hosted Claude Managed Agents in private beta. Companies keep the agent setup, the safe test areas, the login keys, and the full chat history on their own computers. The AI brain itself still answers from a provider the company picks, and Anthropic says it has no remote access to the private server. There is no public price yet, access is only through an account team, and teams must rebuild everything from scratch.
3. Four nonprofits sue Anthropic, Google, Meta, OpenAI, and the SpaceX AI unit in San Francisco. They want a judge to freeze training and releases of any AI stronger than the best systems from November 12 until the companies prove the systems are under human control. The case quotes the leaders own risk warnings. No judge has ruled yet and the claims have not been tested in court.
Would you support a court ordered pause on new AI models, or should companies keep building? Tell us in the comments.
Subscribe to The Daily Neural for a plain English AI news debate every day.
#ainews #artificialintelligence #google #gemini #anthropic #claude #openai #aiagents #technews #aisafety #dailyneural #meta
Watch the video version: https://www.youtube.com/watch?v=45ofisX7oKQ
AI News Today: DeepMind Reads DNA Cell by Cell, ElevenLabs Voice Agents, Hidden AI Sleeper Risk
Google DeepMind can now predict what a DNA change does inside one single type of cell. ElevenLabs wants its voice agents running your calls and meetings. And Microsoft showed an AI can be turned into a sleeper agent by flipping fewer than ten tiny switches.
Alex and Sam break down all three stories in plain English, with every technical term explained the first time it shows up.
Today on The Daily Neural:
1. Google DeepMind releases AlphaGenome 2. Trained on more than 107 million cell profiles, it predicts how DNA changes affect specific cell types instead of whole tissue averages. On one immune cell test it found 30 percent of the variants that change gene activity, nearly double the older model. It is research only and not approved for doctors to diagnose anyone.
2. ElevenLabs makes its big business push. Scribe v3 turns speech into text, Eleven v4 turns text into expressive speech, and Eleven Realtime powers talk back and forth voice agents. The previewed Interaction Engine picks which models run each conversation to balance quality and cost, ElevenAgents is adding Memory for returning customers, and the Partner Program already covers more than 550 partners in 44 countries.
3. Microsoft Research publishes Lazarus. Editing fewer than 10 dormant neurons turned a normal agent model into a sleeper that passed every safety check, then output an attacker command when a hidden trigger appeared. Microsoft says it has not seen anyone use this in the real world, and the strongest defense is comparing a model file against the original untouched version.
Which story matters most to you? AI reading your DNA cell by cell, voice agents that remember you, or a hidden sleeper inside an AI model? Tell us in the comments.
Subscribe to The Daily Neural for a plain English AI news debate every day.
#ainews #artificialintelligence #deepmind #google #alphagenome #elevenlabs #microsoft #aiagents #technews #machinelearning #dailyneural #aisafety
Watch the video version: https://www.youtube.com/watch?v=-G4gjndeHDw
AI News Today: Anthropic $1 Trillion IPO, Claude Fable 5.2, AI Fails Lab Test
Anthropic launched a stock sale that could value the company at about one trillion dollars on the same week it released Claude Fable 5.2. OpenAI then showed its best model scoring only 66.7 percent on a real lab test.
Alex and Sam break down all three stories in plain English, with every technical term explained the first time it shows up.
Today on The Daily Neural:
1. Anthropic releases Claude Fable 5.2 and Mythos 5.2. Same underlying model, two very different doors. Fable is open to paid users, costs 8 dollars per million input tokens after a 20 percent cut, and runs up to 30 percent faster. Mythos has fewer restrictions and stays locked to vetted cybersecurity and life sciences organizations.
2. Anthropic launches its IPO roadshow. The company is offering 825 million shares at 70 to 78 dollars each on Nasdaq under ticker ANT, raising up to about 64 billion dollars, with a separate 26 billion dollar private deal from Google, Microsoft, Nvidia and other giants.
3. OpenAI and Ligo release BioBasicsBench. On 100 routine wet lab tasks the best model, GPT-6 Astra, earned only 66.7 percent partial credit, missed key controls, picked the wrong colony and overcounted colonies by about 12 percent on one plate. OpenAI itself says that score does not mean reliable lab judgment.
Which story matters most to you? The trillion dollar stock debut, the cheaper Claude model, or AI stumbling in a real lab? Tell us in the comments.
Subscribe to The Daily Neural for a plain English AI news debate every day.
#ainews #artificialintelligence #anthropic #claude #openai #ipo #nasdaq #ai #technews #machinelearning #dailyneural #gpt
Watch the video version: https://www.youtube.com/watch?v=Ggd94dZjlUI
AI News Today: Claude Fable 5.2 Gets Cheaper, ChatGPT Hits AWS, Google Builds Secret AI Chips
Anthropic cut the price of its smartest model and locked the strongest version behind a verification check. OpenAI put ChatGPT Enterprise inside Amazon's cloud store. Google built new AI chips designed to keep private data sealed while AI works on it.
Alex and Sam break down all three stories in plain English, with every technical term explained the first time it shows up.
Today on The Daily Neural:
1. Anthropic releases Claude Fable 5.2 and Mythos 5.2. Same underlying model, two levels of access. Fable is open to everyone at 8 dollars per million input tokens, down 20 percent, and up to 30 percent faster. Mythos is restricted to verified cybersecurity and life sciences organizations. Watch out: harder thinking can still cost more per finished task.
2. OpenAI sells ChatGPT Enterprise through AWS Marketplace. Companies can buy it on their existing AWS bill and count eligible spending toward their AWS commitments. No public price, private offers only, and buying through AWS does not by itself keep company data on AWS systems.
3. Google expands its TPU chip lineup with TPU 8tx and TPU 8i Max. Four TPU systems in one year, a first for Google. The 8tx is built for confidential work in banking and government, and the 8i Max doubles memory bandwidth for very long documents. No price and no launch date yet, so every speed claim is Google's own for now.
Which story matters most to you? The cheaper frontier model, ChatGPT at work, or chips that keep data sealed? Tell us in the comments.
Subscribe to The Daily Neural for a plain English AI news debate every day.
#ainews #artificialintelligence #claude #anthropic #openai #chatgpt #aws #google #ai #technews #machinelearning #dailyneural
Watch the video version: https://www.youtube.com/watch?v=7X78ctSHNEE
AI News Today: Mistral Drops a Trillion Parameter Open Model and Shield AI Hits $28.8B
A trillion parameter AI model anyone can download, a 5.3 billion dollar radar deal, and a warning that AI is getting close to mastering biology.
Alex and Sam break down what all three stories mean for regular people, in plain English, with no jargon left unexplained.
Today on The Daily Neural:
1. Mistral Large 4 goes open. Mistral published the weights for a model with about one trillion total parameters. Anyone can download and use it for free, but selling access to it as a service needs a separate deal, and running it takes a serious server, not a home laptop.
2. Shield AI buys Echodyne. The defense company is paying up to 5.3 billion dollars for the Kirkland radar maker, and raising money at a 28.8 billion dollar valuation, more than double its value in March. Its self flying systems are getting better eyes to spot incoming drones.
3. Tencent warns on biology risk. Chief AI scientist Yao Shunyu says AI hacking skills, and soon biology skills, are nearing a point where real harm becomes possible. Tencent says its new Hy4 model was filtered for safety and that outside researchers will get to check its safety report.
Which story worries you most? The open model, the defense deal, or the biology warning? Tell us in the comments.
Subscribe to The Daily Neural for a plain English AI news debate every day.
#ainews #artificialintelligence #mistralai #openai #anthropic #ai #technews #shieldai #tencent #aisafety #machinelearning #dailyneural
Watch the video version: https://www.youtube.com/watch?v=UAh9mXWDa0U
AI News: GPT 6 Cyber Found 948 Linux Exploits and Microsoft Walked Away From Billions
OpenAI built a model that found 948 working Linux exploits in one run. Then Microsoft walked away from tens of billions in extra AI compute. Alex and Sam debate what that means for security, spending, and open models.
In this episode:
1. OpenAI GPT 6 Cyber scores 48.5 percent on ExploitGym, reproduces 1724 Linux kernel flaws, develops working exploits for 948 of them at about 1279 dollars per exploit, and finds 81 high severity Firefox flaws. Access stays gated to approved Daybreak Red customers.
2. Microsoft declines to expand its Nscale Texas data center deal. The existing 43.8 billion dollar work through 2033 stays, Nscale total contract value stays near 103 billion dollars, and Anthropic is in talks for capacity worth nearly 45 billion dollars.
3. Moonshot releases Kimi K3.1 Pro open weights. The 2.8 trillion parameter model has a 1 million token context, ranks first among open weight models on Agent Arena, and needs data center hardware to run.
Subscribe to The Daily Neural for a fast daily breakdown of everything artificial intelligence. Tell us in the comments if you would pay 1279 dollars per exploit for an AI security audit or host a giant open weight model yourself.
#ainews #openai #gpt6 #anthropic #microsoft #kimi #artificialintelligence #cybersecurity #opensource #technews #aiagents #dailyneural
Watch the video version: https://www.youtube.com/watch?v=10NyE72YkxA
AI News Today: Baidu 280B Open Model, Meta Buys Conductor, AI Research Expires
Baidu just released a 280 billion parameter open model as a raw checkpoint, Meta bought the coding agent orchestrator Conductor, and Stanford found 44 percent of famous LLM findings have already expired.
Alex and Sam debate what it all means for developers, researchers and everyone building on AI.
Today on The Daily Neural:
1. Baidu ERNIE 5.5 Flash Base: a text only sparse mixture of experts model with 280 billion total parameters and 14 billion active per token, pre trained on 33.5 trillion tokens with up to 1 million tokens of context, released as a base checkpoint under a custom license that excludes the EU, UK and South Korea.
2. Meta acquires Conductor: the startup that runs multiple AI coding agents in parallel workspaces joins Meta on undisclosed terms, with Conductor Cloud going read only on November 13 and shutting down on November 20.
3. Stanford shelf life study: Owen Queen and James Zou retested 50 widely cited LLM findings on 20 models and found 44 percent expired, 34 percent weakened, and a mean lifespan of just 15.7 months.
Which story matters most? Tell us in the comments.
Subscribe to The Daily Neural for your fast daily breakdown of everything artificial intelligence.
#ainews #artificialintelligence #baidu #ernie #meta #aiagents #codingagents #stanford #llm #opensourceai #technews #thedailyneural
Watch the video version: https://www.youtube.com/watch?v=FBwZzJjBeKU
AI News Today: Tencent Hy4, a $495M AI Kill Switch and Google Schools Lose Search
Tencent just released an open source model deliberately trained to be bad at hacking, CrowdStrike paid a reported 495 million dollars for agent security, and Google told schools that turning off AI means losing Search too.
Alex and Sam debate what it all means for workers, students and everyone deploying AI agents.
Today on The Daily Neural:
1. Tencent Hy4: an open source model trained with cybersecurity data limited to defensive content, with exploit development, reverse engineering and penetration testing filtered out.
2. CrowdStrike agrees to buy Fiddler AI and Dreadnode for a reported 495 million dollars combined and launches an AI Kill Switch in Falcon Guardian that can stop a rogue agent, block relaunches and cut network access.
3. Google flips school AI to off by default. Gemini, Gemini Notebook, Gemini in Classroom and Search stay off until administrators opt in, and schools that reject AI entirely lose Search as well.
Which story matters most? Tell us in the comments.
Subscribe to The Daily Neural for your fast daily breakdown of everything artificial intelligence.
#ainews #artificialintelligence #tencent #hy4 #crowdstrike #aiagents #aisafety #google #gemini #aieducation #technews #thedailyneural
Watch the video version: https://www.youtube.com/watch?v=y8Dfkb5C9s0
AI Edits AI Weights Live, Anthropic $17B Quarter, 400 Hidden Deepfake Apps
An AI just performed surgery on another AI brain and its math score jumped six times. Alex and Sam debate what that really proves, why Anthropic is warning investors that AI could soon speed up its own research, and how 400 innocent looking apps hid deepfake tools in plain sight.
Today on The Daily Neural:
1. MIT CSAIL shows GPT 6 Astra editing Qwen weights directly with no training run, lifting MATH 500 from 7.9 percent to 48.1 percent by changing just 0.016 percent of the model. Repair or real breakthrough? We argue both sides.
2. Anthropic reports about $17 billion in Q3 revenue and $895 billion in future commitments in its amended IPO filing, while investors push its valuation down and new risk language says AI accelerated progress may be imminent.
3. WIRED and AI Forensics expose more than 400 apps posing as pet portrait editors, tutors and workout coaches that used a server side switch to load hidden sexual deepfake tools. Apple removed 19, Google removed most, and the EU ban lands December 2.
Subscribe to The Daily Neural for your fast daily breakdown of everything artificial intelligence. New episodes every weekday.
#ainews #openai #anthropic #claude #gpt #qwen #ai #artificialintelligence #technews #deepfake #mit #aisafety
From the publisher's feed