🗣️ Jeff Bezos shares the advice he received from John Doerr early on at Amazon

“When I first met John Doerr who is the partner at Kleiner Perkins who invested in Amazon, one of the things he said really stuck with me,” Jeff Bezos begins. “What startup companies do is they take their precious early-capital dollars and systematically eliminate risks. That’s what the successful ones do.”

Jeff continues:

“What people often get wrong is that when you’re a startup company, 99% of whether you make it to being a more established company is luck. At Amazon we’ve worked incredibly hard. We’ve cared for our customers. I would put us up against any company in terms of how much we have bled and sweat for our customers. But we had the planets align for us so perfectly in those early days, in terms of the timing and many other things like decisions we made that were poor decisions but turned out to be the right decision anyway and those early days are when the company’s destiny is not really in its own control.”

As a company grows larger and becomes established, Jeff explains, you worry less about externalities.

📊 @tech

☁️ A material that’s lighter than air

Sora Materials unveiled Soramatex at CES 2026, an ultra-light material with a density of 0.0005–0.01 g/cm³, meaning it’s literally lighter than air. Journalists who handled it said it was almost impossible to feel in their hands.

🔸 The company keeps the formula secret, describing it only as “carbon powder,” though its properties resemble graphene aerogel.

🔸 Unlike traditional aerogels, Soramatex is fire-resistant, solving a major safety issue that has limited real-world adoption.

🔸 The material offers thermal insulation, electromagnetic shielding, and sound absorption in a single structure.

🔸 Sora Materials claims Soramatex is recyclable, positioning it as a more sustainable advanced material.

🔸 In aviation, using it as insulation could cut aircraft weight by up to one ton, directly improving fuel efficiency and range.

🔸 Other target use cases include spacecraft, electric vehicles, and “flying cars.”

🔸 Soramatex was developed through a Japanese research program in Aichi Prefecture.

🔸 The company came to CES to find partners for commercialization, signaling a push from lab material to industrial deployment.

If Soramatex scales as promised, weight, not power, may become the next big bottleneck in transportation.

📊 @tech

🧠 How ChatGPT killed stack overflow

Stack Overflow was once the place for programmers to ask questions. At its pandemic peak, it shaped how millions of developers learned to code. But habits have changed and fast.

🔸 Today, it’s often easier to ask a chatbot than to dig through forum threads, comments, and accepted answers.

🔸 As a result, activity has collapsed: the number of new questions last month was roughly the same as in 2008, the year Stack Overflow launched.

🔸 In practical terms, the forum as a living community is fading, fewer questions, fewer answers, less discussion. But the company itself isn’t dead.

🔸 Stack Overflow no longer relies primarily on ads. Its real asset is the massive archive of developer knowledge built over nearly two decades.

🔸 That archive is now monetized by selling access to AI companies, which use the data to train large language models.

Stack Overflow didn’t lose to bad moderation or competition, it lost to a new interface for knowledge itself.

📊 @tech

🔥 A boiler that mines Bitcoin while heating water

The company Superheat has released a household boiler with a built-in Bitcoin miner, it heats water and simultaneously mines BTC.

🔸 The idea is simple: the device consumes the same amount of electricity as a regular boiler, but part of that cost is offset by income from cryptocurrency mining.

🔸 Instead of wasting heat, the miner’s output is fully reused for water heating, improving overall energy efficiency.

🔸 In effect, you’re paying for hot water and getting Bitcoin as a by-product.

🔸 The boiler is priced at $2000, positioning it as a niche product for crypto-friendly households.

It’s less about getting rich on mining and more about turning unavoidable energy use into something marginally profitable.

📊 @tech

💼 ChatGPT wants to help you land your next job

OpenAI is testing a new feature tentatively called ChatGPT Jobs, aimed at job search, career navigation, and professional growth.

🔸 The feature is not publicly available yet, but early details suggest a full-stack assistant for job seekers.

🔸 ChatGPT Jobs will help users improve and tailor their resumes based on role, industry, and seniority level.

🔸 It will identify suitable job openings by matching skills, experience, and career goals not just keywords.

🔸 Users will get interview guidance, including how to answer questions, position strengths, and avoid common mistakes.

🔸 The system is expected to find, compare, and evaluate job offers, helping users weigh compensation, role scope, and growth potential.

🔸 The goal appears to be reducing reliance on traditional job boards and recruiters by turning ChatGPT into a career copilot. No official launch date has been announced.

If this works as intended, job hunting may shift from scrolling listings to having an AI actively represent your interests in the market.

📊 @tech

🔔 Anthropic tightens access to Claude and one change really hurts

There are two pieces of news from Anthropic, both about the same thing: they’re tightening control over how Claude is used for coding, especially Claude Opus 4.5, which many now consider the strongest coding model on the market. One move was expected. The other is painful.

🔸 Anthropic cut off xAI from Claude Opus 4.5 inside Cursor. It turned out xAI engineers were actively using Claude for coding. Anthropic didn’t like that and simply отключили доступ, no ceremony.

🔸 This wasn’t a surprise. Anthropic’s terms explicitly prohibit using their models to build competing AI products. It was only a matter of time. Still, sympathy to the xAI devs.

🔸 The painful part: Anthropic banned using personal Claude subnoscriptions inside third-party clients like OpenCode.

🔸 Previously, OpenCode let users plug in their own paid Claude subnoscription and work with the same modes and limits as the official client but at a much lower cost than buying tokens via the API. That loophole is now closed.

🔸 Anthropic’s official explanation: third-party clients “damage UX.” If something breaks in OpenCode, users might blame Claude itself, hurting Anthropic’s reputation.

🔸 Unofficially, the real issue looks like unit economics and strategic control, not UX. Subnoscription leakage into external tools undermines pricing and platform leverage.

🔸 The result: OpenCode users are scrambling for workarounds on social media but so far, nothing reliable has surfaced.

🔸 From Anthropic’s perspective, the move is understandable. From users’ perspective, it’s brutal.

Claude Opus 4.5 may be the best coding model right now but Anthropic is making it very clear: you’ll use it where and how they want, or not at all.

📊 @tech

We need this in 2026.

📊 @tech

🩺 Anthropic launches Claude for personalized healthcare

Anthropic is entering the health AI space with a new tool designed to help patients and providers make sense of medical data.

🔸 The system, powered by Claude Opus 4.5, can connect to electronic medical records as well as Apple Health and Android Health data.

🔸 It analyzes users’ health metrics, interprets test results, spots correlations between activity and wellbeing, and even suggests questions for doctor visits.

🔸 For healthcare organizations, Claude integrates with Medicare databases, ICD-10 coding, and PubMed, offering research and administrative support.

🔸 Anthropic emphasizes that all data handling complies with US privacy laws. Currently, access is limited to paid subscribers in the United States.

Claude for Healthcare aims to make AI a daily companion for medical insight, bridging personal health data and professional care.

📊 @tech

BREAKING: Anthropic just unveiled "Cowork," a major feature that turns Claude into a fully autonomous virtual assistant for everyone.

It brings the deep agentic capabilities previously reserved for coders to general users, allowing Claude to perform complex tasks directly on your computer.

📊 @tech

🛒 Google unveils a universal protocol to make AI agents real shoppers

Google has introduced the Universal Commerce Protocol (UCP), a new open standard designed to let AI agents handle more of the buying process on behalf of users and businesses.

🔸 Announced at the National Retail Federation (NRF) conference, UCP provides a shared language for AI agents to work across entire commerce workflows from product discovery through checkout and post‑purchase support without bespoke integrations for every retailer or system.

🔸 The protocol was developed in collaboration with major partners including Shopify, Etsy, Wayfair, Target, and Walmart, and is endorsed by more than 20 companies spanning payment networks and retailers.

🔸 UCP is designed to work alongside and interoperate with existing agent‑focused standards such as Agent Payments Protocol (AP2), Agent2Agent (A2A), and the Model Context Protocol (MCP), giving agents flexible interoperability across platforms.

🔸 Google says the protocol will soon power native checkout experiences directly within its AI surfaces including Google Search’s AI Mode and the Gemini app enabling users to complete purchases with Google Pay and shipping info stored in Google Wallet without leaving the conversation.

🔸 PayPal support is planned next, broadening payment options for UCP‑enabled transactions. Retailers can also embed branded “Business Agents” that answer product questions and offer direct discounts inside the AI shopping experience.

🔸 Google’s CEO Sundar Pichai described UCP as a foundational building block for an era where AI interfaces guide users from discovery to purchase seamlessly.

By standardizing how agents talk to merchants, platforms, and payment providers, Google is pushing AI beyond recommendations into actual transaction execution, potentially reshaping digital commerce.

📊 @tech

JUST IN: Google is working on a new tool for Gemini called "Auto Browse". Auto Browse will Chrome into a proper agentic browser, where Gemini will serve as a control interface.

This addition is a continuation of agentic functionalities announced in September.

📊 @tech

🔒 Deep dive guide surfaces for Cursor rival “Antigravity”

A comprehensive walkthrough of Antigravity, a competitor to Cursor and Google’s Windsurf, has appeared online, offering a rare behind-the-scenes look at the platform and its ecosystem of AI agents.

🔸 The guide was created by a developer who has analyzed the full stack, covering everything from installation to monetization.

🔸 It includes detailed instructions on setting up the platform, developing AI-powered features, and even generating revenue using the system.

🔸 The tutorial runs roughly 30 minutes, but it’s dense enough to serve as a complete primer for anyone looking to understand how Antigravity operates.

🔸 The analysis highlights how Antigravity leverages an army of AI agents to automate tasks that rival offerings like Cursor and Google Windsurf.

🔸 For developers and enthusiasts, this guide provides practical, hands-on insights that go beyond marketing materials, showing the platform’s real capabilities.

The post is already circulating among AI communities, suggesting that Antigravity’s architecture may influence future agent-based tools.

📊 @tech

💰 A profitable path to half a million:

1) Rebuild an expensive tool better
2) Launch open-source to attract users + contributors
3) Focus on SEO with programmatic content
4) Build in public for your audience

Iuliia Shnai took Papermark to $500K in 12 months - no funding

📊 @tech

📺 Google is rolling Veo 3.1 updates across Gemini, Flow, AI Studio and APIs.

🔸 Vertical formats support
🔸 Veo 3.1 Ingredients to Video
🔸 Improved ingredients to video consistency
🔸 Upscaling to 1080p and 4K across all Veo models
🔸 Verification of AI-generated videos in Gemini

📊 @tech

💻 AI code generation is turning into a real market

🔸 Analysts estimate the global market at $5.8B in 2025. Forecasts put it at $7.2B this year, with a projected path to $52B by 2035 at a 24.6% CAGR.

🔸 The smoothness of these curves deserves skepticism. Real markets rarely grow that cleanly for a decade.

🔸 Still, AI code generation is already embedded in daily developer workflows. That makes its impact on software labor costs a structural question, not a theoretical one.

📊 @tech

Anthropic puts $1.5M into Python and open source security

🔸 Anthropic signed a 2 year partnership with the Python Software Foundation, committing a total of $1.5M.

🔸 The focus is PyPI security and core infrastructure. One planned move is AI based scanning of every uploaded package. Today, most checks happen only after users report problems.

🔸 The first step is building a large dataset of known malware, with a chance it becomes open source.

This targets the weakest link in modern software supply chains.

📊 @tech