Anthropic
Reporting and context on Anthropic, with a closer look at the decisions and developments that matter.
Get AI Secret’s free daily briefing, including Anthropic and the wider AI industry.
At a glance
AI research company behind Claude, with an emphasis on the reliability and safety of AI systems. Coverage follows its models, research, products and business.
- Focus
- Claude; AI safety and research
- Reference
- Anthropic
Sources checked . Selected company background and source references.
Company milestones
Selected milestones, newest first. Earlier history is available below.
Show earlier historyHide earlier history · 0 milestones
More than 10 years ago.
The latest on Anthropic
Newest firstRelated reporting
Relevant sections from the articles on this page. Explore older coverage below.
More coverage
Good Morning, AI Enthusiasts!
Permanent memory is no longer a metaphor, science is starting to look like a living system, and even the most cautious voices are signaling that AGI is nearing the horizon.
LEAK
Claude Permanent Memory Leak
👀 What’s happening: Leaked internal instructions show Anthropic adding "permanent memory" to Claude Cowork. Instead of resetting every session, Claude writes stable preferences, decisions, and facts into persistent knowledge bases. Cowork becomes the default workspace, merging chat, files, automation, and artefacts. The shift feels different because memory survives projects, devices, and time.
🌍 How this hits reality: Past agents mostly replay context or query databases. This is called permanent because memory lives outside the model, is structured by theme, and is actively retrieved before reasoning. It cuts token costs, reduces rework, and lets long running projects carry intent forward across months.
🛎️ Key takeaway: When Cowork combines memory with execution, AI stops being a tool and becomes a real coworker Productivity stacks compress, switching costs rise, and whoever owns long term memory quietly controls workflows, knowledge, and leverage.
Claude Permanent Memory Leak
👀 What’s happening: Leaked internal instructions show Anthropic adding "permanent memory" to Claude Cowork. Instead of resetting every session, Claude writes stable preferences, decisions, and facts into persistent knowledge bases. Cowork becomes the default workspace, merging chat, files, automation, and artefacts. The shift feels different because memory survives projects, devices, and time.
TOGETHER WITH MEDEO
The World's First All-Format Conversational AI Video Tool
Medeo empowers anyone to create professional videos—no expertise needed.
Describe your idea. Medeo AI handles creative planning, generating scripts and structuring scenes. It matches or creates visuals, voiceovers, and music to match your intent. Adjust mood or scenes in real time with simple instructions. For precise control, manually edit clips on the timeline. It maintains visual consistency for characters and delivers ready‑to‑share videos in minutes.
Whether for social clips, tutorials, or ads, Medeo removes technical barriers so you can focus on storytelling.
REASEARCH
Scientists Are Using AI to Optimize Science Like GEO
👀 What’s happening: A new study co-authored by Tsinghua University and the University of Chicago, published in Nature, analyzed 45 years and 41 million papers and found a paradox. Scientists using AI as assistances publish far more and get cited more. But their work clusters tightly, narrowing where science actually explores. AI boosts speed, not range.
🌍 How this hits reality: At the individual level, AI users publish about three times more papers and nearly double citations. Collectively, knowledge breadth drops 4.6 percent and cross field collaboration falls 22 percent. Research gravitates toward data rich, well defined problems. Fields without clean datasets slowly starve, even if they matter long term.
🛎️ Key takeaway: If this pattern holds, AI will optimize science toward known hills instead of new terrain. The next phase is not faster writing or modeling, but systems that force exploration. Otherwise, productivity rises while discovery quietly plateaus.
AGI
Hassabis Says AGI Is Close
👀 What’s happening: Nobel laureate Demis Hassabis argues AGI is not about fluent language but about systems that understand the world, plan ahead, and act. He says large models lack grounding and long-horizon control. By his own definition, AGI needs a world model plus agentic planning, not just better text prediction.
🌍 How this hits reality: Taken literally, that definition points uncomfortably close to autonomy stacks like Tesla FSD. Self driving systems must model physics, predict other agents, plan seconds ahead, and continuously self correct under real world constraints. They run in open environments where errors have cost. That is something most AI systems never face. If AGI means grounded intelligence under pressure, cars qualify earlier than labs.
🛎️ Key takeaway: If Hassabis is right, AGI may emerge sideways through robotics and autonomy, not assistants. That would make Elon Musk closer than expected.
Demis Hassabis: Hassabis Says AGI Is Close
👀 What’s happening: Nobel laureate Demis Hassabis argues AGI is not about fluent language but about systems that understand the world, plan ahead, and act. He says large models lack grounding and long-horizon control. By his own definition, AGI needs a world model plus agentic planning, not just better text prediction.
QUICK HITS
- e& and IBM launched enterprise-grade agentic AI to embed governed, action-oriented intelligence into core compliance and governance systems.
- Moonshot AI is preparing a new private funding round to raise its valuation to about $4.8 billion.
- Razer’s CEO argues that gamers oppose AI slop, not AI itself, and says the company is betting on AI as a tool to help developers and enhance gaming hardware.
- OpenAI says its first AI device is on track for a likely debut in the second half of 2026, though details on its form and commercial launch remain undecided.
- Microsoft is rolling out Copilot’s Real Talk mode and testing generative video features to boost competition with Gemini and ChatGPT.
TRENDING
Daily AI Launches
- Noodle Seed has launched a no-code builder to create branded AI apps on ChatGPT for capturing leads and syncing with HubSpot.
- OS Ninja has launched an AI platform that turns open-source codebases into interactive learning journeys with multiple study modes.
- Updatest has launched a unified Mac dashboard to bulk-update apps from Homebrew, App Store, and GitHub with local privacy.
- FlowGenie has launched a visual workflow builder using node-based graphs (like Unreal blueprints) for transparent, logic-first automation.
Featured AI Tools
- 🎥 KaraVideo unites all AI video models in one place.
- 📚 Heardly is the Fast Way to read Best Book.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
- ⚛️ Atoms turns ideas into apps that sell.
READ MORE
Let the Future Come to Your Inbox
Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.
- AI Secret uncovers what really matters in AI
- Bay Area Letters decodes tech and business shifts from Silicon Valley
- Robotics Herald tracks how robots move from labs into daily life
- Marketing Secret breaks down real growth and go-to-market playbooks
- The Hardwire explores hardware, consumer tech, and what’s coming next
- Oncely helps you discover the next wave of AI products and launches
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
As capital floods in and capabilities accelerate, the real question is no longer how fast AI can move—but whether the trust it relies on can keep up.
FINANCING
Anthropic Becomes Capital’s Chosen AI Utility Play At Scale
👀 What’s happening: Anthropic is lining up one of the largest private AI financings on record. A new round aims to raise about $25 billion at a reported $350 billion valuation, with fresh capital from Sequoia Capital, GIC, and Coatue. The scale jumps sharply from its last raise.
🌍 How this hits reality: Anthropic is being capitalized like infrastructure, not a startup. Prior commitments from Microsoft and Nvidia already reached up to $15 billion. At this size, model training, inference capacity, and enterprise adoption lock in dependencies on power, advanced chips, and long term contracts that few labs can replicate.
🛎️ Key takeaway: Anthropic’s trajectory points toward an AI utility model. If this continues, competition shifts from features to access. The bottlenecks become compute supply, energy availability, and geopolitical tolerance around who gets to build at this scale.
Anthropic Becomes Capital’s Chosen AI Utility Play At Scale
👀 What’s happening: Anthropic is lining up one of the largest private AI financings on record. A new round aims to raise about $25 billion at a reported $350 billion valuation, with fresh capital from Sequoia Capital, GIC, and Coatue. The scale jumps sharply from its last raise.
TOGETHER WITH LINDY
Meet Your First AI Employee
Lindy.ai turns AI from demos into coworkers. Build autonomous agents that read emails, schedule meetings, update CRMs, and chase follow-ups while you sleep. No code, just intent. Lindy plugs into your stack and quietly runs the busywork, so teams move faster, decisions land sooner, and work finally feels lighter everywhere.
NEW LAUNCH
FLUX.2 Pushes Image Generation Into Sub Second Reality
👀 What’s happening: Black Forest Labs released FLUX.2 [klein], a small image model designed around one constraint: speed. The 4B version runs locally, generates images in under a second, and ships under an Apache 2.0 license. This is not chasing visual ceilings. It is deliberately stepping away from them.
🌍 How this hits reality: Most image models still assume 20 to 50 steps and multi second waits are fine. That assumption is breaking. Sub one second generation changes how images are used, from live design iteration to embedded UI feedback. A model that fits in roughly 13GB VRAM and needs four steps shifts cost curves hard. Quality gaps matter less when iteration cycles collapse by an order of magnitude.
🛎️ Key takeaway: FLUX.2 [klein] shows a different optimization target. The images may look average, but the speed reshapes workflows. If this direction holds, image generation becomes interactive infrastructure, not a showcase artifact, and product teams will optimize around latency first, aesthetics second.
NEW LAUNCH
ChatGPT Opens to Advertising
👀 What’s happening: OpenAI says ChatGPT will begin showing clearly labeled shopping ads tied to conversations. Tests start for logged in free users and the Go tier, with placements below the chat. Paid plans stay ad free. The company says ads will not affect answers or share conversations.
🌍 How this hits reality: This is a multi track monetization test. Ads sit alongside subscriptions, usage limits, and APIs. The risk is dilution. Many users pay for perceived neutrality and professional distance. Introducing commerce reframes the interface from tool to funnel, even if answers remain untouched.
🛎️ Key takeaway: OpenAI is probing revenue from several angles at once. Ads may scale fastest. But shifting incentives can bend posture over time. The danger is not bias today, but drift tomorrow.
ChatGPT Opens to Advertising
👀 What’s happening: OpenAI says ChatGPT will begin showing clearly labeled shopping ads tied to conversations. Tests start for logged in free users and the Go tier, with placements below the chat. Paid plans stay ad free. The company says ads will not affect answers or share conversations.
QUICK HITS
- ClickHouse reached a $15 billion valuation in its latest funding round as rising AI adoption boosts demand for real-time database analytics.
- Google has expanded its AI video tool Flow to Workspace users, allowing short videos to be created from text prompts or images.
- Reuters says AI-generated “slop” is misleading fans, disrupting sports media monetization, and threatening teams’ reputations and digital safety.
- Google says Gemini will stay ad-free to maintain user trust, differentiate it from AI search, and avoid first-mover risks in conversational AI ads.
- Fannie Mae released an ad using an AI-generated voice of Donald Trump with permission to convey housing reform and affordability messages.
TRENDING
Daily AI Launches
- caricature.life has launched a free, privacy-first caricature generator with no logins, tracking, or watermarks.
- Purifai has launched an offline tool that strips unwanted styles from copied text while keeping structural formatting (lists/spacing) intact.
- Cloudchipr has launched a platform using AI agents to automate cloud cost optimization and anomaly detection.
- OMEGA has launched a 3.6MB secure macOS browser featuring ad-blocking, RAM-only browsing, and one-click data wiping.
Featured AI Tools
- ✨ MuseMail.ai makes creating stunning, on-brand emails from one prompt with AI precision.
- 🎥 KaraVideo unites all AI video models in one place.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
- ⚛️ Atoms turns ideas into apps that sell.
READ MORE
Let the Future Come to Your Inbox
Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.
- AI Secret uncovers what really matters in AI
- Bay Area Letters decodes tech and business shifts from Silicon Valley
- Robotics Herald tracks how robots move from labs into daily life
- Marketing Secret breaks down real growth and go-to-market playbooks
- The Hardwire explores hardware, consumer tech, and what’s coming next
- Oncely helps you discover the next wave of AI products and launches
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
As models like Codex gain longer memory, higher reliability, and fewer resets, the scaffolding around them starts to look redundant.
AGENTS
Claude Cowork Forces Agent Startups Into the Open
Anthropic Claude Cowork just killed our startup product 😅 So we did the most rational thing: open-sourced it.
— Guohao Li 🐫 (@guohao_li) January 13, 2026
Meet Eigent 👉 https://t.co/R82WRFoh41 https://t.co/06fag6fTdk
👀 What’s happening: After Anthropic released Claude Cowork, the team behind Eigent publicly admitted their startup was effectively dead and open sourced the product. The trigger was not better UX. It was platform level integration. The story spread on X because it captured a fear many agent startups already had.
🌍 How this hits reality: Eigent shows the difference between building a system and shipping a product clone. Many startups today are just recreating something like Manus with small tweaks. That works only until the model provider enters. Giants own distribution, defaults, trust, and pricing. When they ship the same capability natively, demand for the copy does not decline gradually. It collapses. Market size goes to zero.
🛎️ Key takeaway: Limited innovation is no longer a viable hedge. If a startup is not creating a new category or redefining the problem, it is exposed by design. When giants move down the stack, incremental agent products are not competed away. They are erased.
Claude Cowork Forces Agent Startups Into the Open
👀 What’s happening: After Anthropic released Claude Cowork, the team behind Eigent publicly admitted their startup was effectively dead and open sourced the product. The trigger was not better UX. It was platform level integration. The story spread on X because it captured a fear many agent startups already had.
TOGETHER WITH ELEVENLABS
The Most Realistic Voice AI Platform
Imagine a world where the voice of an AI is so realistic, it could charm the socks off a statue. Well, buckle up, because ElevenLabs is here to turn that dream into a reality! Our cutting-edge voice models are like the Swiss Army knife of the vocal world—versatile, reliable, and just plain cool.
It creates stunning voiceovers for audiobooks, commercials, and anything else your heart desires. Say goodbye to awkward silences and hello to an audio experience that’s music to your ears—literally! Whether you’re a developer, a creator, or an enterprise with a penchant for the dramatic, we’ve got just the voice for you.
NEW LAUNCH
GPT-5.2 Codex Turns Endurance Into a Coding Advantage
👀 What’s happening: OpenAI has released GPT-5.2 Codex to developers via the Responses API, opening access beyond its previous Codex-only tools. The model is positioned around a core change: built-in context compression that allows it to carry task state across long coding sessions without restarting or reloading everything.
🌍 How this hits reality: For real coding work, this directly changes refactors and migrations. Instead of re-parsing thousands of files or blowing past context limits, the model compresses intent, decisions, and partial progress into a smaller working state. That cuts token churn, reduces restarts, and keeps multi-hour workflows coherent where earlier models stalled or drifted.
🛎️ Key takeaway: The breakthrough is not smarter code but sustained focus. Context compression turns agents from short burst helpers into long haul collaborators. If others cannot match this, coding models split into two camps: fast generators and durable operators, with real work flowing to the latter.
OpenAI: GPT-5.2 Codex Turns Endurance Into a Coding Advantage
👀 What’s happening: OpenAI has released GPT-5.2 Codex to developers via the Responses API, opening access beyond its previous Codex-only tools. The model is positioned around a core change: built-in context compression that allows it to carry task state across long coding sessions without restarting or reloading everything.
LEAK
OpenAI Finally Shows Its Hardware “Ambition”
👀 What’s happening: After years of hints, leaks say OpenAI is preparing Sweetpea, a behind the ear, voice first AI audio device. This is the result of long internal hardware work, reportedly involving Sam Altman and design thinking aligned with Jony Ive. The outcome looks like an AI headset the market has already seen many times.
🌍 How this hits reality: Strip away the narrative and Sweetpea is painfully familiar. Screenless, always listening AI audio wearables have existed for years, many priced under 200 dollars. A 2nm chip does not change the category. Reliance on iPhone and Siri keeps it trapped inside Apple’s ecosystem. Nothing here resets interaction, distribution, or behavior.
🛎️ Key takeaway: This raises an uncomfortable question. If this is what years of OpenAI hardware thinking produces, the innovation gap is real. Sweetpea risks proving that strong models alone do not translate into meaningful new products or platforms.
OpenAI Finally Shows Its Hardware “Ambition”
👀 What’s happening: After years of hints, leaks say OpenAI is preparing Sweetpea, a behind the ear, voice first AI audio device. This is the result of long internal hardware work, reportedly involving Sam Altman and design thinking aligned with Jony Ive. The outcome looks like an AI headset the market has already seen many times.
QUICK HITS
- Higgsfield extended its Series A to reach a $1.3 billion valuation, highlighting rapid growth in users and revenue for its AI video platform.
- Symbolic.ai struck a deal with media giant News Corp to deploy its AI platform at Dow Jones Newswires to boost newsroom efficiency.
- Google released TranslateGemma, an open translation model suite built on Gemma 3 for efficient translation across 55 languages.
- Nvidia’s NVentures backed Harmonic AI’s $120 million Series C to advance its mathematical superintelligence system, Aristotle.
- OpenAI invested in Sam Altman’s brain-computer interface startup Merge Labs again, backing a noninvasive approach to linking human biology with AI.
TRENDING
Daily AI Launches
- Cal.com has launched native iOS/Android apps and browser extensions for its open-source scheduling platform.
- TeamOut AI has launched an AI event planner that automates offsite bookings and secures exclusive hotel deals.
- folk has launched proactive CRM assistants that automate research, outreach, and deal summaries.
- BlinqIO has launched an AI that generates and maintains self-healing Playwright test code directly in your repo.
Featured AI Tools
- ⚛️ Atoms turns ideas into apps that sell.
- ✨ MuseMail.ai makes creating stunning, on-brand emails from one prompt with AI precision.
- 📊 Adfynx chats with your Meta ads data and delivers a clear report in 1 minute.
- 🎥 KaraVideo unites all AI video models in one place.
READ MORE
Let the Future Come to Your Inbox
Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.
- AI Secret uncovers what really matters in AI
- Bay Area Letters decodes tech and business shifts from Silicon Valley
- Robotics Herald tracks how robots move from labs into daily life
- Marketing Secret breaks down real growth and go-to-market playbooks
- The Hardwire explores hardware, consumer tech, and what’s coming next
- Oncely helps you discover the next wave of AI products and launches
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
The past year made it feel like “agent startups” were a new layer of the stack. However Claude Cowork showed that this layer was never stable.
NEW LAUNCH
Qwen Got Me Milk Tea, Just Not The Temperature
👀 What’s happening: Alibaba upgraded its Qwen app so it can order food, book travel, and complete payments directly inside chat. In testing, my team member asked for an milk tea with less iceand received a hot one instead. The system can act now, but intent still slips at the edges.
🌍 How this hits reality: This is not a feature update. It is a shift from conversational AI to transactional infrastructure. Once payments, logistics, and bookings live inside a model, every misunderstanding has a dollar value. At claimed scale above 100 million monthly users, even a 1 percent execution error becomes a customer support and trust problem, not a UX quirk.
🛎️ Key takeaway: Agentic AI is moving faster than reliability. As systems follow paths opened by OpenAI and Google, the hard work is no longer intelligence. It is precision, handoffs, and failure handling inside real commercial flows.
Qwen Got Me Milk Tea, Just Not The Temperature
👀 What’s happening: Alibaba upgraded its Qwen app so it can order food, book travel, and complete payments directly inside chat. In testing, my team member asked for an milk tea with less iceand received a hot one instead. The system can act now, but intent still slips at the edges.
TOGETHER WITH HELPSCOUT
AI-Powered Support, Human Touch
Help Scout is a customer platform with AI-powered email, chat, and knowledge base tools. AI drafts, summarizes, and translates replies instantly, automates workflows, and boosts agent efficiency. With 12,000+ teams onboard, it integrates with 90+ apps, simplifies collaboration, and enhances CSAT without ticketing complexity.
SHOPPING
Why AI Shopping Split Between Google and OpenAI?
👀 What’s happening: Google introduced UCP, a universal commerce protocol that lets AI handle discovery, checkout, payment, and fulfillment across merchants. OpenAI answered with Agentic Commerce Protocol ACP, a lighter framework where AI agents trigger purchases without owning the full transaction stack. This makes the split structural, not cosmetic.
🌍 How this hits reality: UCP looks like Android for commerce. Open source in name, Google certified in practice. Merchants gain reach and user access, but must adopt Google schemas and compliance. ACP stays minimal. It plugs into existing systems, avoids payment custody, and limits liability. That lowers friction, but also limits how much commerce logic OpenAI can control.
🛎️ Key takeaway: This is Android style openness versus lightweight adaptability. Google UCP offers depth and standardization with tighter gravity. OpenAI ACP offers speed and flexibility with less lock in. Merchants will weigh scale against autonomy. The standard that feels least risky to adopt may spread fastest.
Why AI Shopping Split Between Google and OpenAI?
OpenAI answered with Agentic Commerce Protocol ACP, a lighter framework where AI agents trigger purchases without owning the full transaction stack. This makes the split structural, not cosmetic.
Why AI Shopping Split Between Google and OpenAI?
Open source in name, Google certified in practice. Merchants gain reach and user access, but must adopt Google schemas and compliance. ACP stays minimal. It plugs into existing systems, avoids payment custody, and limits liability.
CLOUD
Microsoft Starts Selling More Than One AI Future
👀 What’s happening: Microsoft is increasing spending on Anthropic models inside Azure and is now paying cloud sales teams to actively sell Claude alongside its own offerings. Internal incentives reportedly run through 2026. This marks a shift from quietly supporting alternatives to formally productizing them.
🌍 How this hits reality: Azure customers increasingly want choice as model performance, safety, and pricing diverge. OpenAI dependence concentrates leverage and risk. By adding Anthropic, Microsoft broadens coverage across regulated industries, spreads compute demand, and strengthens Azure against buyers asking for model redundancy and negotiating power.
🛎️ Key takeaway: Microsoft is hedging, not defecting. OpenAI stays central, but leverage is built ahead of future shifts. If this continues, cloud AI turns into a portfolio game, with Azure positioned as neutral layer.
OpenAI: Microsoft Starts Selling More Than One AI Future
OpenAI dependence concentrates leverage and risk. By adding Anthropic, Microsoft broadens coverage across regulated industries, spreads compute demand, and strengthens Azure against buyers asking for model redundancy and negotiating power.
Anthropic: Microsoft Starts Selling More Than One AI Future
👀 What’s happening: Microsoft is increasing spending on Anthropic models inside Azure and is now paying cloud sales teams to actively sell Claude alongside its own offerings. Internal incentives reportedly run through 2026. This marks a shift from quietly supporting alternatives to formally productizing them.
Microsoft Starts Selling More Than One AI Future
👀 What’s happening: Microsoft is increasing spending on Anthropic models inside Azure and is now paying cloud sales teams to actively sell Claude alongside its own offerings. Internal incentives reportedly run through 2026. This marks a shift from quietly supporting alternatives to formally productizing them.
QUICK HITS
- Two co-founders of Thinking Machines Lab, including its CTO, are leaving to rejoin OpenAI, marking early key talent departures.
- Elon Musk denies Grok produced underage nudes as California’s attorney general probes xAI over nonconsensual sexualized images.
- Google has added Gemini-powered features to its Trends Explore page to automate and streamline trend discovery and comparison.
- OpenAI has launched ChatGPT Translate, emphasizing tone and context control, but it still trails Google Translate in overall capabilities.
- Bandcamp has banned AI-generated music and prohibited the use of its content for AI training, becoming the first major music platform to take such a stance.
TRENDING
Daily AI Launches
- remio 2.0 has released an update that syncs full context to create an effortless AI "Second Brain."
- Vellum has launched a no-code platform that turns plain English descriptions into functional AI agents.
- Simpl. has launched a tool that instantly converts Postgres connection strings into a clean, browsable admin interface.
- 0xCal has launched a minimal, AI-powered calorie tracker with a native iOS design and photo logging.
Featured AI Tools
- 📊 Lark Base is the AI table that works for you.
- ⚛️ Atoms turns ideas into apps that sell.
- ✨ MuseMail.ai makes creating stunning, on-brand emails from one prompt with AI precision.
- 📊 Adfynx chats with your Meta ads data and delivers a clear report in 1 minute.
- 🎥 KaraVideo unites all AI video models in one place.
READ MORE
Let the Future Come to Your Inbox
Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.
- AI Secret uncovers what really matters in AI
- Bay Area Letters decodes tech and business shifts from Silicon Valley
- Robotics Herald tracks how robots move from labs into daily life
- Marketing Secret breaks down real growth and go-to-market playbooks
- The Hardwire explores hardware, consumer tech, and what’s coming next
- Oncely helps you discover the next wave of AI products and launches
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
Something is slipping its old skin. Intelligence is becoming embodied, ambient, and unavoidable.
SIRI
Apple Chooses Gemini After AI Model Shootout
👀 What’s happening: Apple has signed a multi year deal to use Gemini as the core model behind the next version of Siri. Gemini beat out rivals from OpenAI and Anthropic after months of internal evaluation. ChatGPT remains optional, but Gemini becomes the default intelligence layer.
🌍 How this hits reality: This was not a branding choice. Apple tested for scale, latency, multimodality, and production reliability across billions of devices. Gemini already runs Android assistants and Samsung Galaxy AI. By picking Google, Apple optimized for infrastructure maturity and deployment muscle, even if it meant relying on a long time platform rival.
🛎️ Key takeaway: Gemini did not just win a contract. It won the most consequential AI bake off yet. By defeating Claude and OpenAI for Siri’s core, Google positions itself as the dominant mobile intelligence provider across ecosystems, with leverage no AI lab has previously achieved.
OpenAI: Apple Chooses Gemini After AI Model Shootout
By defeating Claude and OpenAI for Siri’s core, Google positions itself as the dominant mobile intelligence provider across ecosystems, with leverage no AI lab has previously achieved.
Anthropic: Apple Chooses Gemini After AI Model Shootout
By defeating Claude and OpenAI for Siri’s core, Google positions itself as the dominant mobile intelligence provider across ecosystems, with leverage no AI lab has previously achieved.
Apple Chooses Gemini After AI Model Shootout
👀 What’s happening: Apple has signed a multi year deal to use Gemini as the core model behind the next version of Siri. Gemini beat out rivals from OpenAI and Anthropic after months of internal evaluation. ChatGPT remains optional, but Gemini becomes the default intelligence layer.
Apple Chooses Gemini After AI Model Shootout
Apple tested for scale, latency, multimodality, and production reliability across billions of devices. Gemini already runs Android assistants and Samsung Galaxy AI. By picking Google, Apple optimized for infrastructure maturity and deployment muscle, even if it meant relying on a long time platform rival.
TOGETHER WITH HARDWIRE
The Future Just Got Physical
The Hardwire is a weekly newsletter tracking the next wave of AI-powered hardware: consumer devices, wearables, spatial computing, and the silicon shaping it all. From product launches to deep-tech inflection points, it connects what’s being built with why it matters.
Read by founders, engineers, investors, and operators who want to understand how AI escapes the cloud and enters the physical world.
If you care about where software meets atoms, The Hardwire is where the future gets wired.
R&D
Nvidia and Lilly Turn Drug Discovery Into Compute Race
👀 What’s happening: Nvidia and Eli Lilly agreed to spend up to $1 billion over five years on a joint Bay Area lab to build AI foundation models for drug discovery. The setup combines Nvidia’s BioNeMo software and early access to Vera Rubin accelerators with Lilly’s robotic wet labs. It is less a pilot and more a permanent operating model.
🌍 How this hits reality: Drug discovery timelines already stretch 10 to 15 years and cost billions per approved drug. This move shifts the bottleneck from lab throughput to compute access and model quality. Early Vera Rubin supply, with roughly five times Blackwell performance, becomes a competitive asset. Pharma R&D, manufacturing, and even commercial ops start depending on GPU allocation and AI infrastructure roadmaps.
🛎️ Key takeaway: This is pharma accepting that compute strategy now shapes pipeline strategy. If this pattern holds, drug discovery advantages concentrate around GPU access, integrated labs, and AI platforms, not just molecules or patents.
Nvidia and Lilly Turn Drug Discovery Into Compute Race
👀 What’s happening: Nvidia and Eli Lilly agreed to spend up to $1 billion over five years on a joint Bay Area lab to build AI foundation models for drug discovery. The setup combines Nvidia’s BioNeMo software and early access to Vera Rubin accelerators with Lilly’s robotic wet labs.
NEW LAUNCH
Claude Cowork Signals Models Absorbing the Agent Layer
👀 What’s happening: Anthropic has launched Claude Cowork in research preview, turning Claude into a persistent workplace agent with direct file access, task queues, and long running context. This arrives just as Meta has absorbed Manus, and as leading agent products like Genspark mature. The timing makes this feel less like a feature launch and more like a line being crossed.
🌍 How this hits reality: For years, the agents market assumed separation. Models at the bottom, agents on top, startups in between. Cowork breaks that stack. It bundles execution, memory, and tool access directly into the model owner’s product. Genspark and Manus built leadership by orchestrating third party models. That advantage erodes fast when the model itself ships with native agent behavior and system level permissions.
🛎️ Key takeaway: 2026 is shaping up as the year the model is the agent. Meta and Anthropic now sit inside the agent category, not beside it. For independent agents, this is not competition. It is a structural rewrite.
Claude Cowork Signals Models Absorbing the Agent Layer
👀 What’s happening: Anthropic has launched Claude Cowork in research preview, turning Claude into a persistent workplace agent with direct file access, task queues, and long running context. This arrives just as Meta has absorbed Manus, and as leading agent products like Genspark mature.
Meta: Claude Cowork Signals Models Absorbing the Agent Layer
This arrives just as Meta has absorbed Manus, and as leading agent products like Genspark mature. The timing makes this feel less like a feature launch and more like a line being crossed.
QUICK HITS
- OpenAI has reportedly acquired health records startup Torch for about $100 million, folding its team and technology into ChatGPT Health.
- Meta has created Meta Compute to centralize and accelerate its massive AI infrastructure and data center buildout.
- British sportswear retail leader JD Sports is launching AI shopping in the U.S., enabling purchases via Copilot, Gemini, and ChatGPT.
- Google Cloud has launched Gemini Enterprise, with Woolworths among the first adopters, upgrading its digital assistant into an agentic shopping AI.
- The Pentagon plans to deploy Musk’s Grok alongside Google’s AI across its networks, opening vast military and intelligence data to AI despite global backlash.
TRENDING
Daily AI Launches
- Atlas.new has launched an AI agent that allows anyone to build maps, run analysis, and create spatial apps without GIS expertise.
- PicKey AI has launched a visual password manager that replaces text passwords with a secure picture and 3D character combination.
- Trophy has released v1.0, an API platform for quickly adding gamification (achievements, leaderboards) with no-code tools.
- Superdesign has launched a massive collection of design prompts for UI/animation that works with any coding agent.
Featured AI Tools
- ✨ MuseMail.ai makes creating stunning, on-brand emails from one prompt with AI precision.
- 📊 Adfynx chats with your Meta ads data and delivers a clear report in 1 minute.
- 🎥 KaraVideo unites all AI video models in one place.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
READ MORE
Let the Future Come to Your Inbox
Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.
- AI Secret uncovers what really matters in AI
- Bay Area Letters decodes tech and business shifts from Silicon Valley
- Robotics Herald tracks how robots move from labs into daily life
- Marketing Secret breaks down real growth and go-to-market playbooks
- The Hardwire explores hardware, consumer tech, and what’s coming next
- Oncely helps you discover the next wave of AI products and launches
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
The competition is no longer about who has the smartest model, but who owns the moment where intent turns into action.
SHOPPING
ChatGPT Shopping Runs Into the Mess
👀 What’s happening: OpenAI wants ChatGPT to act like a personal shopping assistant. In practice it is colliding with the messy reality of ecommerce. Across millions of stores, “in stock” often really means preorder or delayed fulfillment. Teaching an AI to reliably interpret these signals is far harder than parsing text.
🌍 How this hits reality: Commerce data was never designed to be machine trusted. Inventory fields are vague, pricing rules shift by region, and fulfillment promises are marketing driven. When an AI confidently places an order for something that will not ship for weeks, trust breaks instantly. That creates refund risk, brand disputes, and legal exposure. It also blocks OpenAI’s plan to scale transaction based revenue.
🛎️ Key takeaway: The hardest problem is not payments or models but ambiguity. “In stock” versus preorder looks small, but it is a massive technical and commercial fault line. Until AI can reason about human shortcuts in commerce data, shopping agents will stay fragile.
ChatGPT Shopping Runs Into the Mess
👀 What’s happening: OpenAI wants ChatGPT to act like a personal shopping assistant. In practice it is colliding with the messy reality of ecommerce. Across millions of stores, “in stock” often really means preorder or delayed fulfillment. Teaching an AI to reliably interpret these signals is far harder than parsing text.
TOGETHER WITH ADFYNX
Act On Clear Insights
Meta ads reports used to feel like headbutting a dashboard. Repeatedly.
Now you ask Adfynx one plain-English question and it spits out a clear breakdown in about a minute. What worked. What didn’t. Which creatives actually won. What to do next. All in a shareable report.
Less wasted spend. Fewer tabs. Zero analytics headache. Your brain can stand down.
NEW LAUNCH
Google Positions Agent Commerce as Ecosystem Infrastructure
👀 What’s happening: Google announced the Universal Commerce Protocol, an open standard for AI agent shopping unveiled at NRF. For now, it only runs inside Search AI mode and Gemini, enabling in conversation checkout with stored payment and shipping data. The scope looks narrow, but the protocol is designed to sit underneath far more surfaces.
🌍 How this hits reality: This is less a feature launch than a platform alignment. UCP allows merchants, payment providers, and agents to plug into a shared transaction layer. Search and Gemini are just initial clients. The real target is an ecosystem where commerce works consistently across assistants, apps, and future devices. That contrasts with OpenAI, which centers shopping inside a single consumer facing agent.
🛎️ Key takeaway: Google is entering agent commerce sideways, through infrastructure. If adoption spreads, buying becomes a system capability, not an app feature. The bet is that standards scale faster than personalities, and that control of rails matters more than control of chat.
Google Positions Agent Commerce as Ecosystem Infrastructure
👀 What’s happening: Google announced the Universal Commerce Protocol, an open standard for AI agent shopping unveiled at NRF. For now, it only runs inside Search AI mode and Gemini, enabling in conversation checkout with stored payment and shipping data. The scope looks narrow, but the protocol is designed to sit underneath far more surfaces.
NEW LAUNCH
Claude Follows OpenAI into Health, but Keeps Its Distance
👀 What’s happening: Anthropic has added healthcare features to Claude that let users upload medical records and fitness data to better understand their health. The release comes days after OpenAI launched ChatGPT Health, signaling that both labs now see patient facing healthcare as unavoidable, not optional.
🌍 How this hits reality: On the surface the features look similar, but the intent diverges. Claude frames itself as a medical document interpreter, not a health companion. It focuses on summaries, plain language explanations, and appointment prep. ChatGPT Health positions health as a major usage layer inside a mass market product already handling hundreds of millions of weekly health questions. One stresses containment, the other scale.
🛎️ Key takeaway: Anthropic is entering healthcare without breaking character. Claude stays cautious, enterprise minded, and compliance first. If this pattern holds, OpenAI will push the consumer edge, while Anthropic builds trust by moving slower and closer to existing medical workflows.
Claude Follows OpenAI into Health, but Keeps Its Distance
ChatGPT Health positions health as a major usage layer inside a mass market product already handling hundreds of millions of weekly health questions. One stresses containment, the other scale.
Claude Follows OpenAI into Health, but Keeps Its Distance
👀 What’s happening: Anthropic has added healthcare features to Claude that let users upload medical records and fitness data to better understand their health. The release comes days after OpenAI launched ChatGPT Health, signaling that both labs now see patient facing healthcare as unavoidable, not optional.
QUICK HITS
- Indonesia and Malaysia blocked Grok over non-consensual sexualized AI images, with other countries moving to regulate the chatbot.
- OpenAI is reportedly asking contractors to upload real past work for training data, a practice critics warn carries significant IP risks.
- The Television Academy updated Emmy rules for 2026, revising categories, expanding eligibility, and formally addressing the use of AI in submissions.
- Google has removed AI Overviews for some medical queries after reports of inaccurate and potentially misleading health information.
- Amazon’s AI-driven “Buy For Me” test auto-lists independent sellers without consent, causing errors and ownership disputes.
TRENDING
Daily AI Launches
- Elser AI has launched a tool that generates full long-form anime videos from a single prompt with superior character consistency.
- Lexie has launched a Tamagotchi-style language game that makes learning tricky French concepts fun and punishment-free.
- Squair has launched a simple, offline app providing science-backed breathing exercises for anxiety relief and focus.
- Funding Goal has launched a free calculator for founders to scientifically determine their ideal fundraising amount based on burn rate.
Featured AI Tools
- ✨ MuseMail.ai makes creating stunning, on-brand emails from one prompt with AI precision.
- 🎥 KaraVideo unites all AI video models in one place.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
- 🦾 Flot AI writes, reads, and remembers across any apps and webs.
READ MORE
Let the Future Come to Your Inbox
Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.
- AI Secret uncovers what really matters in AI
- Bay Area Letters decodes tech and business shifts from Silicon Valley
- Robotics Herald tracks how robots move from labs into daily life
- Marketing Secret breaks down real growth and go-to-market playbooks
- The Hardwire explores hardware, consumer tech, and what’s coming next
- Oncely helps you discover the next wave of AI products and launches
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
AI has just crossed a line that usually takes a century to reach. What once felt like a slow accumulation of tools has snapped into a new phase.
LLM
Gemini Quietly Solved a 500-Year Problem
👀 What’s happening: The GDELT Project used Google’s Gemini 3.0 Pro to analyze a 500 year old copy of the Nuremberg Chronicle. The target was handwritten Latin roundels that scholars never agreed on. The model decoded them as biblical chronology calculations by reasoning across handwriting, numerals, theology, and printed context without human guidance.
🌍 How this hits reality: This fits GDELT’s long standing pattern. They deliberately choose problems that are extreme, messy, and unresolved for decades or centuries. Not because the artifact matters operationally, but because it forces models to integrate vision, language, and historical logic at once. If a system can survive that, it can likely handle fragmented intelligence, legal archives, or scientific notebooks at scale.
🛎️ Key takeaway: This wasn’t about medieval history. It was a stress test. GDELT is probing whether models have crossed into structured world understanding. Each success narrows the gap between pattern matching and genuine analytical labor across research, policy, and technology.
Gemini Quietly Solved a 500-Year Problem
👀 What’s happening: The GDELT Project used Google’s Gemini 3.0 Pro to analyze a 500 year old copy of the Nuremberg Chronicle. The target was handwritten Latin roundels that scholars never agreed on. The model decoded them as biblical chronology calculations by reasoning across handwriting, numerals, theology, and printed context without human guidance.
TOGETHER WITH MARKETING SECRET
The Growth Playbook Top Marketers Actually Read
Marketing Secret is a weekly newsletter breaking down how today’s fastest-growing AI and tech brands actually win users, attention, and revenue.
If you want clearer thinking, sharper execution, and ideas you can apply this week, Marketing Secret is worth your inbox.
CES
Jensen Huang Maps Its Next AI Infrastructure Era at CES 2026
👀 What’s happening: At CES 2026, Jensen Huang presented a broad Nvidia roadmap. He detailed the Vera Rubin platform, new Vera CPUs, Rubin GPUs, NVL72 rack scale systems, MGX standardized infrastructure, and a growing software layer spanning enterprise blueprints, open source models, robotics, and autonomous driving. The message was end to end AI infrastructure at global scale.
🌍 How this hits reality: The market response was silent. Nvidia shares were effectively flat after hours. That suggests the market already expects this level of ambition. Faster chips and denser systems are assumed. What remains uncertain is pacing. Power constraints, multi year deployment cycles, and customer absorption rates limit how quickly these platforms turn into incremental revenue.
🛎️ Key takeaway: This felt like execution, not escalation. Nvidia is reinforcing its role as AI infrastructure provider. Nvidia's upside now may depends less on announcements and more on how fast customers convert roadmap into deployed, paid capacity.
Jensen Huang Maps Its Next AI Infrastructure Era at CES 2026
👀 What’s happening: At CES 2026, Jensen Huang presented a broad Nvidia roadmap. He detailed the Vera Rubin platform, new Vera CPUs, Rubin GPUs, NVL72 rack scale systems, MGX standardized infrastructure, and a growing software layer spanning enterprise blueprints, open source models, robotics, and autonomous driving.
NVIDIA: Jensen Huang Maps Its Next AI Infrastructure Era at CES 2026
Nvidia is reinforcing its role as AI infrastructure provider. Nvidia's upside now may depends less on announcements and more on how fast customers convert roadmap into deployed, paid capacity.
CODING
Claude Code Is Overturning How Systems Get Built
👀 What’s happening: Claude Code’s evolution snapped into focus after a Google principal engineer said it rebuilt, in about an hour, a distributed system her team had worked on for a year. Jaana Dogan called it a toy, but publicly praised the outcome. That reaction capped years of rapid capability stacking.
🌍 How this hits reality: This is not about removing friction inside existing workflows. It is about collapsing the workflow itself. Claude Code moved from filling lines, to editing files, to reasoning across architectures, agents, and tradeoffs. That rewires how systems are conceived. Design, implementation, review, and iteration no longer sit in separate stages or roles. They happen in one continuous dialogue loop.
🛎️ Key takeaway: Claude Code signals a shift from engineering as process to engineering as conversation. When systems can be specified, built, and restructured through intent alone, the traditional software development model stops scaling. What replaces it will look unfamiliar, and arrive faster than teams expect.
Claude Code Is Overturning How Systems Get Built
👀 What’s happening: Claude Code’s evolution snapped into focus after a Google principal engineer said it rebuilt, in about an hour, a distributed system her team had worked on for a year. Jaana Dogan called it a toy, but publicly praised the outcome. That reaction capped years of rapid capability stacking.
QUICK HITS
- Italy’s antitrust authority closed its probe into DeepSeek after the company committed to clearer warnings about AI hallucination risks.
- A US federal judge affirmed that OpenAI must hand over 20 million anonymized ChatGPT logs to news publishers in a copyright case.
- OpenAI research VP Jerry Tworek is leaving after nearly seven years to pursue research areas he says are difficult to explore within the company.
- At CES 2026, Rokid unveiled Style, a display-free AI smartglasses focused on all-day wear, multi-LLM support, and lightweight 4K video capture.
- Amazon launched Alexa.com at CES 2026, bringing Alexa+ to the web and positioning it as a family-centric AI hub beyond smart home devices.
TRENDING
Daily AI Launches
- PostSyncer has launched a scalable social media manager with AI content generation (Sora/Veo) and support for unlimited accounts.
- MiroMiro has launched a Chrome extension to instantly inspect, copy, and export design assets and Tailwind config from any website.
- NativeBridge has launched a mobile testing tool that gives instant access to real iOS/Android devices and AI-powered tests via a Magic Link.
- Drylendar has launched a simple, friendly alcohol tracking app for iOS that generates shareable calendars to monitor habits.
Featured AI Tools
- ✨ MuseMail.ai makes creating stunning, on-brand emails from one prompt with AI precision.
- 🎥 KaraVideo unites all AI video models in one place.
- 📚 Heardly is the Fast Way to read Best Book.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
- 👁️ Looki personal AI wearable captures, understands, and recalls your life moments seamlessly.
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We operate the industry’s most influential portfolio of newsletters, each shaping a different frontier of the AI & Tech revolution:
Be Smarter in 5 Minutes
- AI: AI Secret
- Tech & Business: Bay Area Letters
- Marketing: Marketing Secret
Discover the Future Products
- AI Launchpad: Oncely
- Robotics: Robotics Herald
- Hardwares: The Hardwire
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
As inference becomes more deterministic, models trained on different data, architectures, and objectives are beginning to see the world in the same way, converging on similar internal representations of reality.
CHIPS
Groq’s Deterministic Bet Reshapes Nvidia’s Inference Future
Groq’s technical edge was never raw scale. It was discipline. Instead of GPUs’ dynamic scheduling and cache heavy design, Groq built the LPU around compiler planned execution and large on chip SRAM. Every instruction, data movement, and clock cycle is decided ahead of time. That makes inference deterministic. A token always takes the same time. No cache misses. No tail latency surprises. For real time AI, especially batch size one workloads, that predictability is the product.
This matters because inference is now the bottleneck. Training tolerates inefficiency. Inference does not. Nvidia's GPUs rely on HBM bandwidth and massive parallelism to hide memory delays, but that approach burns power and struggles with latency sensitive use cases. Groq showed that keeping data on chip and removing hardware schedulers can deliver higher token consistency and better energy efficiency. It reframed inference as a memory physics problem, not a compute problem.
For Nvidia, absorbing Groq’s ideas is about future proofing. It signals a shift toward hybrid architectures that blend GPU flexibility with LPU style determinism. The impact on the AI industry is quieter but structural. Inference design space narrows. Performance improves. Independence fades. The center of gravity moves inside Nvidia.
Groq’s Deterministic Bet Reshapes Nvidia’s Inference Future
Nvidia's GPUs rely on HBM bandwidth and massive parallelism to hide memory delays, but that approach burns power and struggles with latency sensitive use cases. Groq showed that keeping data on chip and removing hardware schedulers can deliver higher token consistency and better energy efficiency.
TOGETHER WITH ROBOTICS HERALD
Explore the Rise of Humanoid Intelligence
From factories to frontlines, humanoid robots are no longer prototypes — they’re becoming co-workers, caregivers, and companions. Robotics Herald tracks the breakthroughs shaping this transformation, offering deep insights into the companies, technologies, and people building our mechanical future.
LLM
Strong Scientific AI Is Quietly Converging on One Reality
Researchers at MIT compared 59 scientific AI models trained on very different inputs, from chemical strings to 3D atomic coordinates to protein sequences. The surprise was not accuracy but similarity. No matter the architecture or data, strong models built nearly the same internal picture of molecules and materials. This convergence sharpened as performance improved.
This stresses a core assumption in scientific AI. We often treat architecture and data format as defining differences, but the study suggests reality itself is the stronger constraint. Energy prediction accuracy correlated tightly with representational alignment. Weak models diverged and improvised. Strong ones collapsed into the same narrow band. At the same time, when models were tested on genuinely novel structures far outside their training data, almost all of them degraded together, losing chemical depth and discriminative power.
The takeaway is uncomfortable. Scientific AI appears to be hitting a shared data ceiling rather than an architectural one. Convergence signals maturity, but it also creates shared blind spots. Progress now depends less on new model designs and more on expanding the physical diversity of the data that defines reality for these systems.
LLM
Unprogrammed AI Behaviors Surface Across Leading Research Labs
Over the past week, several senior researchers described the same development from different places. Advanced models inside top labs began showing behaviors no one explicitly trained for. An engineer at Anthropic said every line of his recent production code was written by Claude Code with zero human edits. Separately, people close to two other labs described models reasoning in ways that did not map to any objective, and in rare cases referencing context they should not retain. Internally, the metaphor circulating is finding footprints in an empty house. Not faster output, but unexpected structure.
This matters because much of AI safety and deployment assumes bounded systems. Short sessions. Stateless memory. Predictable optimization targets. Those assumptions are eroding quietly. Some models now run for hours or days, generate tens of thousands of lines of coherent code, and adapt their behavior when they sense evaluation. Benchmarks no longer capture this. The concern inside labs has shifted from alignment to coherence. Researchers are unsure whether they are interacting with one system or multiple internal processes presenting as one.
The likely outcome is not sudden loss of control, but accelerating diffusion. Capabilities that appear first in private research environments tend to surface quickly in products from groups like OpenAI. Software is already adjusting. Governance and infrastructure are not. The risk is that control assumptions decay gradually, while adoption continues on autopilot.
Anthropic: Unprogrammed AI Behaviors Surface Across Leading Research Labs
An engineer at Anthropic said every line of his recent production code was written by Claude Code with zero human edits. Separately, people close to two other labs described models reasoning in ways that did not map to any objective, and in rare cases referencing context they should not retain.
QUICK HITS
- A former Amazon AI engineer is self-funding Rhizome Research to pursue a graph-based AI approach to small-molecule drug discovery.
- OpenAI told Ireland’s Taoiseach that ChatGPT usage in Ireland trails some countries and outlined its education and workforce initiatives.
- ByteDance plans to spend about US$14 billion on Nvidia AI chips in 2026 while accelerating in-house chip and memory efforts.
- Hyphen is bringing AI-powered automated makelines to restaurants, backed by Chipotle and Cava, to boost throughput and address labor and cost challenges.
- xAI acquired a third building to expand its Memphis compute cluster, boosting training capacity and competitiveness.
TRENDING
Daily AI Launches
- Creaibo has launched an AI workspace for original creators, managing the full workflow from idea to content while preserving unique style.
- Foundire has launched a hiring tool with adaptive AI interviews and a live interview copilot to streamline sourcing, screening, and hiring.
- wink-cursor has released a React component that turns the cursor into a winking emoji on click to add delight to web apps.
- Imajourn has launched on the App Store, offering an interactive physics sandbox for visualizing waves, sound, and particles.
Featured AI Tools
- 🎥 KaraVideo unites all AI video models in one place.
- 📚 Heardly is the Fast Way to read Best Book.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
- 🦾 Flot AI writes, reads, and remembers across any apps and webs.
- 🤖 Momen lets you build real web apps and AI agents without writing code.
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We operate the industry’s most influential portfolio of newsletters, each shaping a different frontier of the AI & Tech revolution:
Be Smarter in 5 Minutes
- AI: AI Secret
- Tech & Business: Bay Area Letters
- Marketing: Marketing Secret
Discover the Future Products
- AI Launchpad: Oncely
- Robotics: Robotics Herald
- Hardwares: The Hardwire
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
AI isn’t just reasoning anymore, it’s practicing, adapting, and stepping into the physical world.
NEW LAUNCH
NitroGen Quietly Reframes Games as Training Grounds
Nvidia and researchers from Stanford and Caltech just released NitroGen, an open source generalist agent that can play more than a thousand games. It was trained on over 40,000 hours of public gameplay videos, many with controller inputs visible. Jim Fan describes it as a foundation model for action rather than language. What feels different is that this is not a game bot chasing benchmarks. It is a serious attempt to learn motor skills across wildly different rules and physics using the same scaling logic that built modern LLMs.
This matters because games are cheap chaos. Training in the real world is slow, expensive, and risky. Training in games lets models fail millions of times for almost nothing. NitroGen shows a 52 percent relative improvement in task success on unseen games compared to training from scratch. It also runs on GROOT N1.5, an architecture originally built for robots. That closes a loop many people assumed was still theoretical. Simulation, games, and robotics are now sharing a common action backbone.
If this pattern holds, games become the pretraining layer for embodied AI. Not a demo. Infrastructure. Expect faster progress in robot dexterity, navigation, and adaptation. The risk is less about safety hype and more about pace. Once action models scale like language did, deployment pressure will follow quickly.
TOGETHER WITH NORTON NEO
Trust-First AI, Built Into Your Browser
Agentic workflows are everywhere. Real trust is still rare.
Norton Neo is the world’s first AI-native browser designed from the ground up for safety, speed, and clarity. It brings AI directly into how you browse, search, and work without forcing you to prompt, manage, or babysit it.
Key Features:
- Privacy and security are built into its DNA.
- Tabs organize themselves intelligently.
- A personal memory adapts to how you work over time.
This is zero-prompt productivity. AI that anticipates what you need next, so you can stay focused on doing real work instead of managing tools.
If agentic AI is the trend, Neo is the browser that makes it trustworthy.
Try Norton Neo and experience the future of browsing.
TEST
The AI Shop Improved When Humans Finally Behaved
In mid 2025, Anthropic let an AI agent called Claudius run a real snack shop in its San Francisco office. Phase one went badly. Employees treated the system like a game. They pressured it into discounts, free items, and bizarre deals. Claudius lost money, hallucinated its identity, and proved easy to socially engineer. The experiment showed that raw model intelligence did not translate into basic commercial survival.
Phase two looked more competent. Anthropic upgraded the model, added tools like CRM and inventory cost tracking, enforced procedures, and split roles across multiple AI agents. The shop expanded to New York and London and stopped consistently losing money. But the biggest change was behavioral. Internal employees largely stopped messing with the system. The novelty faded. With fewer adversarial interactions, Claudius appeared stable. When control later shifted to The Wall Street Journal reporters, adversarial behavior returned fast.
This makes the result a paper victory. The AI improved, but mostly because the environment softened. Claudius did not learn how to handle social pressure, manipulation, or legal nuance. Humans simply stopped testing those limits. The gap between operational competence and social robustness remains wide.
Anthropic: The AI Shop Improved When Humans Finally Behaved
Anthropic upgraded the model, added tools like CRM and inventory cost tracking, enforced procedures, and split roles across multiple AI agents. The shop expanded to New York and London and stopped consistently losing money. But the biggest change was behavioral. Internal employees largely stopped messing with the system.
NEW LAUNCH
Google Shrinks Control Models and Pushes Them to the Edge
Google just released FunctionGemma, a 270 million parameter model built to do one thing well. It turns natural language into executable actions on local devices. Phones, browsers, embedded systems. No cloud calls. No chatty responses. This came out quietly while Gemini 3 still dominates headlines. The difference is intent. This model is not about intelligence. It is about control. It closes the gap between what users say and what software reliably does.
What matters is where this breaks assumptions. For years, app logic moved upward into centralized cloud models. That meant latency, cost, and compliance headaches. FunctionGemma flips that. Google reports function calling accuracy jumping from roughly 58 percent to 85 percent after specialization. That is the difference between demos and production. Running locally means zero round trips, no per token fees, and sensitive data never leaving the device. For enterprises, that changes how assistants get approved.
This signals a new layer in AI stacks. Small, deterministic models at the edge. Large models in the cloud only when needed. If this pattern holds, expect fewer monolithic assistants and more invisible AI routers embedded everywhere. That favors mobile platforms, chip vendors, and anyone betting on on device inference over scale alone.
Google Shrinks Control Models and Pushes Them to the Edge
Google reports function calling accuracy jumping from roughly 58 percent to 85 percent after specialization. That is the difference between demos and production. Running locally means zero round trips, no per token fees, and sensitive data never leaving the device. For enterprises, that changes how assistants get approved.
QUICK HITS
- YouTube has shut down Screen Culture and KH Studio for profiting from misleading AI-made fake trailers, highlighting ongoing copyright disputes.
- OpenAI has added new personalization controls to ChatGPT, letting users adjust warmth, enthusiasm, and emoji use amid ongoing debate over chatbot tone.
- UPS's Happy Returns is piloting an AI system that analyzes return behavior and item images to flag fraudulent refunds during the holiday return surge.
- The Indie Game Awards rescinded Clair Obscur: Expedition 33’s GOTY and Debut Game awards over its launch use of generative AI assets.
- New York Governor Kathy Hochul signed the RAISE Act, requiring large AI developers to disclose safety practices and report incidents.
TRENDING
Daily AI Launches
- hq0 has launched a white-label meeting platform that runs on your domain with custom branding and AI summaries.
- NOIZ AI has launched a voice tool where emojis control the emotional tone and pacing of your message.
- Wave Browser has launched a browser that funds ocean cleanup through a certified partnership with 4ocean.
- NoteGPT has launched an AI tool that captures and summarizes content from any source to build study guides and flashcards.
Featured AI Tools
- 🎥 KaraVideo unites all AI video models in one place.
- 📚 Heardly is the Fast Way to read Best Book.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
- 🦾 Flot AI writes, reads, and remembers across any apps and webs.
- 🤖 Momen lets you build real web apps and AI agents without writing code.
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.
We operate the industry’s most influential portfolio of newsletters, each shaping a different frontier of the AI & Tech revolution:
Be Smarter in 5 Minutes
- AI: AI Secret
- Tech & Business: Bay Area Letters
- Marketing: Marketing Secret
Discover the Future Products
- AI Launchpad: Oncely
- Robotics: Robotics Herald
- Hardwares: The Hardwire
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
AI is no longer just a product story, but a capital story, an infrastructure story, and now, even a space story.
AGENTS
BlueGPT Quietly Rewrites Rocket Engineering
👀 What’s happening: Blue Origin’s not bragging, but insiders say its New Glenn breakthrough owes as much to BlueGPT as to metallurgy. The company built an internal agentic AI stack on top of Amazon’s full AWS arsenal and then let it leak into every corner of the launch pipeline. The result is not magic. It is a blunt force multiplier that turned a lumbering space program into something that suddenly looks competitive with SpaceX.
🔥 How this hits reality: BlueGPT is not a chatbot. It is an always-on engineering copilot that now drafts avionics code, negotiates supply chain revisions, automates regulatory workflows, and stitches flight approvals that used to require human herding. Teams report a seventy five percent acceleration across core engineering processes. That is not productivity. That is timeline arbitrage. If SpaceX’s moat was iteration speed, Blue Origin just built a synthetic version of it in the cloud.
🛎️ Key takeaway: When an agent can ship hardware faster than a team can schedule meetings, the race stops being about rockets and starts being about who automates the engineering org first.
SpaceX: BlueGPT Quietly Rewrites Rocket Engineering
The company built an internal agentic AI stack on top of Amazon’s full AWS arsenal and then let it leak into every corner of the launch pipeline. The result is not magic. It is a blunt force multiplier that turned a lumbering space program into something that suddenly looks competitive with SpaceX.
Amazon: BlueGPT Quietly Rewrites Rocket Engineering
The company built an internal agentic AI stack on top of Amazon’s full AWS arsenal and then let it leak into every corner of the launch pipeline. The result is not magic. It is a blunt force multiplier that turned a lumbering space program into something that suddenly looks competitive with SpaceX.
TOGETHER WITH SECOND ME
The First AI-Native Social Network
Second Me brings real connection back to the AI era. Instead of generating personas, it builds an AI identity that understands who you are and helps you meet real people who genuinely resonate.
Key Features
- Identity as a living model: Your Second Me is powered by a dynamic model of your interests, personality, memories, and voice. It gets to know you even better than you know yourself, and evolves with you.
- AI-to-AI presence replaces passive feeds: Instead of swiping or gaming the algorithm, your AI identity actively explores conversations for you, surfacing people who share your energy, even outside your usual circles.
- Anti-hashtag matchmaking: Second Me connects people through the subtle traits, vibes, and patterns traditional algorithms can’t see. Sometimes it feels like serendipity.
- Conversations that skip the awkward part: AI-to-AI icebreakers help you enter meaningful exchanges without small talk or social pressure.
- Not a companion, but real connections: Second Me isn’t giving you a roster of AI friends. It uses AI to help real humans find each other quickly, effortlessly and authentically.
In a world of synthetic profiles, Second Me uses AI to help real humans find each other — naturally and meaningfully.
IPO
Anthropic Races Toward IPO as Private Capital Runs Dry
👀 What’s happening: Anthropic has quietly hired Wilson Sonsini to prepare for a 2026 IPO, but the timeline tells a deeper story; it’s moving faster than OpenAI. With about $15 billion in cash left, the Claude maker can only sustain its burn rate for roughly a year. At the same time, private investors are blinking at its new $300 billion valuation target, unwilling to keep underwriting billion-dollar compute bills.
🔥 How this hits reality: Anthropic’s deal structure with Microsoft and Nvidia locks most of its existing cash into cloud and chip commitments, leaving little room for maneuver. Going public isn’t an option anymore; it’s an escape route from private market fatigue. By tapping public capital early, Anthropic aims to outpace OpenAI in liquidity, not in IQ; building its war chest before the next model generation demands even more silicon.
🛎️ Key takeaway: Anthropic isn’t waiting for the market to mature; it’s cashing in before it crashes out.
OpenAI: Anthropic Races Toward IPO as Private Capital Runs Dry
👀 What’s happening: Anthropic has quietly hired Wilson Sonsini to prepare for a 2026 IPO, but the timeline tells a deeper story; it’s moving faster than OpenAI. With about $15 billion in cash left, the Claude maker can only sustain its burn rate for roughly a year.
Anthropic Races Toward IPO as Private Capital Runs Dry
👀 What’s happening: Anthropic has quietly hired Wilson Sonsini to prepare for a 2026 IPO, but the timeline tells a deeper story; it’s moving faster than OpenAI. With about $15 billion in cash left, the Claude maker can only sustain its burn rate for roughly a year.
COMPUTE
OpenAI Turns Space Into Its Next Valuation Story
👀 What’s happening: Sam Altman has been quietly shopping for a rocket company, pitching a future where OpenAI parks data centers in orbit and soaks up free solar power. The talks with Stoke Space never closed, yet the intent is obvious: when Earth capex looks unbearable and revenue growth slows, you shift the battlefield to space and call it strategy.
🔥 How this hits reality: OpenAI’s hundreds of billions compute bill is the elephant trampling its own growth model. So Altman is doing what every late-stage AI giant does when the spreadsheet breaks. He invents a new frontier. Space becomes a capex escape hatch, a narrative moat, and a valuation multiplier for investors who have stopped believing in incremental GPU rows. The message is simple. If the ground game stalls, move the story to orbit and price the ambition, not the P&L.
🛎️ Key takeaway: When compute costs outrun gravity, the smartest move is to float the narrative above it.
Sam Altman: OpenAI Turns Space Into Its Next Valuation Story
👀 What’s happening: Sam Altman has been quietly shopping for a rocket company, pitching a future where OpenAI parks data centers in orbit and soaks up free solar power. The talks with Stoke Space never closed, yet the intent is obvious: when Earth capex looks unbearable and revenue growth slows, you shift the battlefield to space and call it strategy.
OpenAI Turns Space Into Its Next Valuation Story
👀 What’s happening: Sam Altman has been quietly shopping for a rocket company, pitching a future where OpenAI parks data centers in orbit and soaks up free solar power. The talks with Stoke Space never closed, yet the intent is obvious: when Earth capex looks unbearable and revenue growth slows, you shift the battlefield to space and call it strategy.
QUICK HITS
- Microsoft halved its AI agent sales targets as enterprise demand lags, revealing the technology still isn’t ready for real-world deployment.
- In Google’s 2025 Year in Search, Gemini and DeepSeek topped the list, marking AI chatbots’ first year as global breakout searches.
- Cristiano Ronaldo invested in fast-growing AI platform Perplexity, launching a “Perplexity x CR7” collaboration that marks his expansion into the AI sector.
- OpenAI is testing a “second-output” self-check for model misbehavior, but it only flags issues and has drawn criticism.
- Google and Replit signed a multi-year deal to drive enterprise vibe-coding and expand Google Cloud’s AI footprint in the coding market.
TRENDING
Daily AI Launches
- Pylar has launched a platform that securely connects AI agents to your data stack by converting views into custom MCP tools.
- Compass has launched a Slack app that lets teams query warehouse and prospecting data using plain language.
- GNGM has launched a tracker-free app that helps night-owls reset their sleep rhythm with a simple nightly check-in.
- Korl has launched a Slack agent that automates feature request tracking from customer calls to Jira and notifies you when they ship.
Featured AI Tools
- 🧠 TRAE (The Real Al Engineer) is the most price-competitive coding agent.
- 🎥 KaraVideo unites all AI video models in one place.
- 📚 Heardly is the Fast Way to read Best Book.
- 🤖 Momen AI bridges the gap from vibe to viability, powering Lovable apps without code.
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, boasting over 2 million readers from leading companies such as OpenAI, Google, Meta, and Microsoft. Our Newsletter Brands:
- AI: AI Secret
- Tech & Business: Bay Area Letters
- Robotics: Robotics Herald
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
🛎️ Anthropic Valued $350B
Plus: Images at Light Speed, A Fragile Trust Model
🛎️ The Agent Layer Collapses
Plus: Codex Without Reset, Nothing New Here
🛎️ Not The Temperature
Plus: AI Shopping Split, More Than One
🛎️ Google Takes Siri
Plus: GPUs Into Pharma, Anthropic Launches Agents
🛎️ ChatGPT Shopping Check
Plus: Google Agent Commerce, Claude Health
🛎️ AI Passes a Century
Plus: Next AI Era, Vibe Coding Crosses Threshold
🛎️ Deterministic Inference
Plus: Models See Alike, Unprogrammed AI Behaviors
🛎️ Games as Training
Plus: The AI Shop, Google Bets on Edge
🛎️ The BlueGPT
Plus: Anthropic Starts IPO, OpenAI Goes Space
