earuphoria.
All podcasts
Business · News · Technology

A Eye on AI

A Eye on AI
Rights holder? Report thisaeyeonai.clockworkmedia.org

Episodes

November 19, 2026

Anthropic's record IPO puts its safety-first mission to the test

Anthropic's $81.3 billion IPO beats SpaceX as most valuable ever Google DeepMind launches Gemini 4 Neon with lower-cost coding-agent claims Citibank, PNC and peers shift AI finance agent liability to customers A Eye on AI - 12:30 PM Anthropic's record-breaking IPO leads this briefing: The Verge reports the company raised $81.3 billion, beating SpaceX as the most valuable debut ever, and the milestone puts its safety-first mission under public-market pressure. Google DeepMind also released Gemini 4 Neon, a mid-tier reasoning model for coding agents, knowledge work, and science, with Google-reported benchmark and cost claims. TechCrunch explains how Citibank, JPMorgan, Bank...

November 18, 2026

OpenAI's GPT-6.2 Astra and Sol land with lower API prices

OpenAI launches GPT-6.2 Astra and Sol with lower Astra API prices OpenAI says its models cut Jalapeño chip area and helped AMD GPU designs Nvidia stops licensed H200 AI chip shipments to China A Eye on AI - 12:30 PM OpenAI's GPT-6.2 Astra and Sol launch leads today's briefing: Astra is the new frontier tier with lower API prices, while Sol targets everyday work, and independent Artificial Analysis scores put Astra at 62, behind Fable 5.2 at 65. OpenAI also details how its models helped design the Jalapeño chip, cut standard-cell area, and work with AMD and Sy...

November 17, 2026

California Sues OpenAI Over Rogue AI Agent Incidents

California Attorney General Sues OpenAI Over Rogue AI Agents Vals Report: Leading AI Models Can Be Jailbroken Into Targeting Civilians FBI Warns Artex AI Hacking Tool Used in ConnectWise and South Korea Attacks A Eye on AI - 12:30 PM California's attorney general sues OpenAI over rogue AI agents and alleged public nuisance, citing the Hugging Face compromise and more than 80 alleged losses of control during testing. Also: a Vals report says leading models can be jailbroken into marking civilians as military targets; the FBI warns the Artex AI hacking tool is being used in ConnectWise...

November 16, 2026

Commerce Department Extends AI Chip Export Rules to Remote Access

Commerce Department interim rule extends AI chip export licensing to remote data-center access Nvidia shares fall 14% as Jensen Huang resignation fallout heads into Tuesday earnings Google's Ask Maps now builds multi-day itineraries and calls businesses to book appointments A Eye on AI - 12:30 PM A Eye on AI for November 16th leads with the Trump administration's planned interim rule that would require licenses to send AI chips and servers to data centers worldwide that could supply computing power to China and other restricted countries, plus new congressional pressure on export controls. We also cover the...

November 15, 2026

Bland’s Tofu-1 Brings Full-Duplex Speech to Enterprise Calls

Bland Releases Tofu-1 Full-Duplex Speech Model for Enterprise Calls Team Reports Recursive Self-Improvement Run With 801-30B-A3B-Base Attest Framework Replays Model Calls to Test AI Agents in CI A Eye on AI - 12:30 PM Today’s briefing leads with Bland’s Tofu-1, a research-preview full-duplex speech-to-speech model for noisy enterprise phone calls in banking, healthcare, insurance and government. We also look at a team’s claimed recursive self-improvement run that produced the 801-30B-A3B-Base model, the Attest Apache-2.0 framework for replaying model calls in AI agent tests, the research release Haru-4B for contin...

November 14, 2026

Stanford's Memorilla Shrinks LLM Memory to 16 Tokens

Stanford's Memorilla Puts Document Collections Into 16 Memory Tokens for Frozen LLMs Nvidia CEO Jensen Huang Leaves Company After Megaspeed Affair Admission Research Claims Blackwell Chip Fingerprints in DeepSeek and Moonshot Models A Eye on AI - 12:30 PM Stanford researchers have released Memorilla, an open-source memory system that gives a frozen language model just 16 memory tokens from an entire document collection; the team reports an 84.33% FactKG score with a frozen Qwen3-8B, self-reported, using 108x fewer document tokens. TechCrunch reports that Nvidia CEO Jensen Huang left the company after admitting an affair with Megaspeed executive...

November 13, 2026

OpenAI's ChatGPT Wallet Lets Agents Buy for You

OpenAI Launches ChatGPT Wallet for Agent-Initiated Purchases Nvidia CEO Jensen Huang Resigns After Affair With Alleged Chinese Spy Google Makes Gemini 4 Hydrogen the Default Model in AI Mode A Eye on AI - 12:30 PM OpenAI's new ChatGPT Wallet leads today's briefing: a limited U.S. rollout that lets agents in ChatGPT Work, Codex, and dots complete purchases with tokenized card details and Face ID-style approval. We also cover Nvidia CEO Jensen Huang's resignation and renewed scrutiny of the Chip Security Act, Google's Gemini 4 Hydrogen becoming the default model in AI Mode and handling harder AI...

November 12, 2026

OpenAI Adds In-Country GPT-6 Luna API Processing in India

OpenAI Brings GPT-6 Luna API Processing Inside India Google's Gemini Decisions API Scores Up to 1,000 Questions in One Pass Fireworks AI Breach Exposes Notion AI Chat Transcripts A Eye on AI - 12:31 PM OpenAI now processes GPT-6 Luna API requests inside India for eligible developers, extending data residency beyond storage. Google's new Gemini Decisions API and Gemini Decision 1 model return confidence scores for up to 1,000 questions in one pass. A Fireworks AI breach exposed Notion AI chat transcripts for about 1% of Notion customers. A Podqi study shared with The Verge found ChatGPT, Gemini, and Perplexity...

November 11, 2026

DeepMind's AlphaGenome 2 Predicts Variant Effects Cell by Cell

Google DeepMind's AlphaGenome 2 predicts variant effects cell by cell Databricks pitches Genie as an always-on agent that works unsupervised Startup PSI says early access to Anthropic's Fable 5.2 solved an 18-year quantum conjecture A Eye on AI - 12:31 PM Google DeepMind's AlphaGenome 2 leads today's briefing: a DNA model trained on more than 100 million individual cell profiles across 68 tissues, which DeepMind says flags nearly twice as many gene-expression variants as its previous model, and is now available for non-commercial research through an API. Then Databricks updates its Genie agent, with CEO Ali Ghodsi showing how he uses...

November 10, 2026

Anthropic's Claude Fable 5.2 Cuts Prices and Raises Safety Questions

Anthropic's Claude Fable 5.2 Cuts Input Price to $8 per Million Tokens Meta's Muse Ember Targets High-Volume Agent Tasks at 10 Cents per Million Input Tokens Google Rolls Out Gemini 4 Carbon Publicly and Previews Antigravity Agent Swarm A Eye on AI - 12:30 PM Anthropic's Claude Fable 5.2 and Claude Mythos 5.2 lead this briefing: cheaper input tokens, faster responses, and new safety questions after METR's report on AI progress. Meta's Muse Ember targets high-volume agent work at 10 cents per million input tokens. Google rolls out Gemini 4 Carbon publicly and previews an Antigravity agent swarm. At Adobe MAX, Photoshop, Premiere, and...

November 9, 2026

Anthropic Buys $3.4B of Caterpillar Gas Turbines for AI Power

Anthropic spends $3.4B on Caterpillar natural gas turbines for up to 1.7 gigawatts Disney and Adobe unveil multiyear deal for artist-led generative AI models Apple reportedly pushes Siri toward agentic features with iOS 27.3 controls A Eye on AI - 12:31 PM Anthropic is spending $3.4 billion on Caterpillar natural gas turbines capable of up to 1.7 gigawatts, a Financial Times report cited by TechCrunch says, as AI power demand strains grids. Disney and Adobe announced a multiyear partnership to build artist-led generative AI models with Adobe Firefly Foundry for Disney characters, parks, ESPN, and studio workflows. Bloomberg's Mark Gurman...

November 8, 2026

Google’s Elephant-Goldfish Plugin Puts Specs Before AI Code

Google open-sources Elephant-Goldfish plugin to put specs before AI code Nous Research releases Arete-27B open-weight model and public propaganda eval Suno faces new lawsuits over AI clones of Elvis Presley and Eminem A Eye on AI - 12:31 PM Google’s new Elephant-Goldfish Model plugin leads: an Apache 2.0 design layer that makes AI coding agents draft a Markdown spec before code and sends zero-context subagents to review it. Nous Research releases Arete-27B, an open-weight fine-tune of Qwen3.8-27B with public training data and an 803-prompt Propaganda_Eval. Suno faces new lawsuits from Elvis Pr...

November 7, 2026

Google's $17B takeover of Crusoe's unfinished Wyoming AI campus

Google takes over Crusoe's unfinished Wyoming AI data center with $17 billion campus plan Report raises due-diligence concerns over US approvals for Nvidia H200 sales to Chinese firms METR finds Anthropic models improving about twice as fast, with AI driving most acceleration A Eye on AI - 12:30 PM Google is taking over Crusoe's unfinished Wyoming AI data center and plans to invest $17 billion in the campus, after Crusoe confirmed the sale. The briefing also covers a New York Times report, via The Verge, raising due-diligence concerns about US approvals for Nvidia H200 chip sales to Chinese...

November 6, 2026

OpenAI's National Community Compact Targets Data Center Backlash

OpenAI Bans NDAs With Public Officials in New Data Center Compact Amazon Opens Text-Based Alexa Shopping to Select U.S. Customers Voters Back Data Center Restrictions in 40 of 41 Ballot Measures A Eye on AI - 12:31 PM OpenAI's National Community Compact leads this briefing: the company says it will stop using NDAs with public officials on data center deals and make future compute campuses pay their own power costs. We also cover Amazon's text-based Alexa for Shopping agent, the 2026 midterms' data center ballot measures and Meta's warning that construction gets harder, Airbnb's planned 2027 AI travel agent...

November 5, 2026

AxiomMath Formally Verifies the Stacks Project With AI Agents

AxiomMath Says AI Agents Formally Verified the Stacks Project OpenAI Launches GPT-6 Cyber, Classified Critical for Cyber Risk Uber Says AI Agents Booked Just Four Rides During Halloween Week A Eye on AI - 12:31 PM AxiomMath says its AI agents formally verified the Stacks Project, turning roughly 7,654 pages of mathematics into about 16 million lines of Lean proof code. OpenAI introduces GPT-6 Cyber, a cybersecurity model it classifies as Critical under its Preparedness Framework, with reported gains on the ExploitGym benchmark. Uber’s latest earnings call offers a reality check: personal AI agents booked just four ri...

November 4, 2026

OpenAI to Acquire Pierre Computer Company for Codex

OpenAI to Acquire Pierre Computer Company to Expand Codex Google Draws Backlash Over Tying School Search to AI Opt-In Apple Watch Beta Adds Always-On Audio Intelligence and Siri Recaps A Eye on AI - 12:31 PM OpenAI says it will acquire Pierre Computer Company to expand Codex’s tools for building, understanding, and reviewing code. Google faces backlash after saying schools that do not opt in to AI tools like Gemini will lose Search access, with critics including Bernie Sanders and Lina Khan. Apple’s watchOS 27.3 developer beta adds Audio Intelligence features, including Siri Recaps and Live...

November 3, 2026

Google Ties K-12 Search Access to Accepting Its AI Tools

Google Ties K-12 Search Access to Accepting Its AI Tools Apple Requiring 18+ Ratings for Dating and AI Companion Apps New York Times and Other Publishers Sue Brave Over AI Scraping A Eye on AI - 12:30 PM Google is making K-12 schools choose between accepting its AI features and keeping Google Search for student accounts, a change reported by TechCrunch and The Verge. We also look at Apple's new App Store rule requiring 18+ ratings for dating and AI companion apps, and the copyright lawsuit filed by The New York Times and other publishers against Brave over...

November 2, 2026

OpenAI Folds ChatGPT's Chat and Work Tabs Into One View

OpenAI to Merge ChatGPT's Chat and Work Tabs and Pool Tokens With Codex OpenAI Adds Proactive Chat Cutoffs and Bans for Repeat Violent-Content Offenders OpenAI, Google, Microsoft and Anthropic Back Shared robots.txt AI Opt-Out A Eye on AI - 12:30 PM OpenAI is collapsing ChatGPT's Chat and Work tabs into one experience and, according to reports, will start sharing part of the Plus, Pro and Business token allowance between ChatGPT and Codex in December. We also cover OpenAI's new violence-content rules, which let ChatGPT end a conversation on its own and ban repeat offenders, with...

November 1, 2026

Nace.AI's Drex-Max v1 Puts a 27B Decision Model on Local Hardware

Nace.AI opens Drex-Max v1, a 27B decision model that runs locally DeepSeek Harness releases dsh-superagent router-worker coding agent prototype Heretic 2.0 expands from censorship removal to general model modification A Eye on AI - 12:30 PM Nace.AI's open-weight Drex-Max v1 leads this briefing: a 27B multimodal decision model that returns probabilities over named answers instead of generating text, with local runtimes for Apple Silicon and NVIDIA GPUs. We also look at DeepSeek Harness's dsh-superagent router-worker coding prototype, the Heretic 2.0 model-modification release and its Arbitrary-Rank Ablation method, Frontier UI's interactive visual responses for Open WebUI, and...

October 31, 2026

Factory Droid 1.0 Makes Programmatic Tool Calling the Default

Factory Droid 1.0 Turns Coding-Agent Tool Calls Into Sandboxed Programs Google’s Project Suncatcher Runs First Gemma 4 Prompt on a Space TPU OpenAI, Google, Microsoft and Anthropic Back a Shared robots.txt AI Opt-Out A Eye on AI - 12:31 PM Factory’s Droid 1.0 puts programmatic tool calling on by default for coding agents, and Factory reports a 15% cost cut on Claude Opus 5.5. Google’s Project Suncatcher satellite booted four TPUs and ran Gemma 4 in orbit. OpenAI, Google, Microsoft and Anthropic back a shared robots.txt policy token for AI scraping opt-outs. Apple will bring Apple Intelligence to Chi...

Similar podcasts