{"$schema_note":"Generated by /api/feed from Turso. Latest approved day only. Do not hand-edit.","updated":"2026-07-25","issue_no":74,"topic":"Claude Opus 5 Launch","supporting_topic":"AMD Helios challenges Nvidia","tldr":"Claude Opus 5 launched at Opus 4.8's price while more than doubling its top score, and AMD's Helios rack answered by pulling a 2-gigawatt Anthropic order to challenge Nvidia on the training floor. Running a strong model keeps getting cheaper even as the frontier climbs, with chip startup Etched reaching a $10.3 billion valuation on an inference-only design. Office-work agents kept coming, from Cognition folding a messaging assistant into its Devin coder to a Hoffman-backed lab training models on how clerks click through their day. Monday.com cut a fifth of its staff the same week it rebuilt around those agents.","cover":{"url":"/api/cover/2026-07-25?v=2026-07-25%2002%3A21%3A44","alt":"A lone figure climbs a steep circuit-etched staircase while a price tag stays pinned at the base, suggesting AI power rising as cost holds flat."},"weekly":{"week_start":"2026-07-20","week_end":"2026-07-26","headline":"The model got cheaper this week. Your lock-in moved.","subheadline":"Open weights and a flat-priced Opus 5 turned the model into the swappable part; the week's real bets sat one layer up and one layer down.","url":"/weekly/2026-07-20"},"recent_digests":[{"date":"2026-07-25","topic":"Claude Opus 5 Launch","lead_title":"Claude Opus 5: Anthropic doubles the score and keeps the price flat","supporting_topic":"AMD Helios challenges Nvidia","tl_dr":"Claude Opus 5 launched at Opus 4.8's price while more than doubling its top score, and AMD's Helios rack answered by pulling a 2-gigawatt Anthropic order to challenge Nvidia on the training floor. Running a strong model keeps getting cheaper even as the frontier climbs, with chip startup Etched reaching a $10.3 billion valuation on an inference-only design. Office-work agents kept coming, from Cognition folding a messaging assistant into its Devin coder to a Hoffman-backed lab training models on how clerks click through their day. Monday.com cut a fifth of its staff the same week it rebuilt around those agents.","issue_no":74,"cover":{"url":"/api/cover/2026-07-25?v=2026-07-25%2002%3A21%3A44","alt":"A lone figure climbs a steep circuit-etched staircase while a price tag stays pinned at the base, suggesting AI power rising as cost holds flat."}},{"date":"2026-07-24","topic":"AI Agent Tooling Ships","lead_title":"AI model theft: White House accuses China's Moonshot of copying Anthropic's Fable","supporting_topic":"cheap open-weight models","tl_dr":"AI agent tooling shipped in force while cheap open-weight models kept undercutting the majors. Amazon's Bedrock AgentCore went generally available as a declarative harness, LangGraph 1.0 made Model Context Protocol tools first-class nodes, and OpenAI's Presence put governed voice agents inside BBVA and SoftBank. On the model side, the US Energy Department and Arcee AI opened Genesis-Science-1, Poolside shipped a leaner 118-billion-parameter coder, and a new tokenizer called Gigatoken ran a thousand times faster than Hugging Face. In Washington, the White House accused China's Moonshot of distilling Anthropic's Fable, and Treasury floated sanctions.","issue_no":73,"cover":{"url":"/api/cover/2026-07-24?v=2026-07-24%2002%3A24%3A55","alt":"A figure traces a machine's inner wiring while an official points across the room and cheap copies roll out a low-priced gate, suggesting a fight over copied AI."}},{"date":"2026-07-21","topic":"China's Open Model Surge","lead_title":"Alibaba Qwen 3.8: a 2.4T open-weight model it calls second only to Fable 5","supporting_topic":"Anthropic's free Fable 5 window closes","tl_dr":"China's open models are reshaping which systems developers can actually run: Alibaba previewed Qwen 3.8, a 2.4-trillion-parameter open-weight model it calls second only to Anthropic's Fable 5, days after Moonshot's Kimi K3 hit its GPU ceiling and froze new signups. Anthropic's free Fable 5 window closed on July 19, so teams that leaned on it now pay $10 and $50 per million tokens or switch. Brussels ordered Google to open Android to rival AI assistants and share search data. The money kept moving too: Databricks touched a $188 billion valuation, Tempus bought cancer-test maker Personalis for $1.5 billion, and CuspAI raised $450 million to industrialize materials discovery.","issue_no":72,"cover":{"url":"/api/cover/2026-07-21?v=2026-07-21%2002%3A23%3A48","alt":"A crowd pours through an open gate scattering pages while a coin turnstile shuts beneath a low ceiling of chips, as open models surge against cost and compute limits."}}],"items":[{"id":"2026-07-25--cognition-acquires-poke-devin","title":"Cognition buys Poke to give its Devin coding agent a personality","summary":"For a backend engineer, this hints at where coding tools are heading: less command line, more colleague. If your team is picking an AI pair-programmer, weigh how it communicates and asks for context, not only its benchmark score. The felt experience of working with the thing is about to become a real differentiator.","analysis":"The pitch is that how an agent talks is now worth paying for. Cognition, the company behind the Devin coding agent, acquired The Interaction Company of California, maker of the messaging assistant Poke, in a deal it put in the low nine figures. Poke lives inside iMessage, WhatsApp and Telegram and handled more than 100 million messages in three months by chatting through scheduling and travel tasks. Cognition wants that conversational feel wired into Devin so the tool reads like a co-worker who can take a joke rather than a terminal that spits diffs. The models are commoditizing, so the interface becomes the moat.","url":"https://techcrunch.com/2026/07/24/why-cognition-bought-poke-ai-personality-is-becoming-a-competitive-advantage/","source":"TechCrunch","date":"2026-07-25","tags":["agents"]},{"id":"2026-07-25--claude-opus-5-flat-price","title":"Claude Opus 5: Anthropic doubles the score and keeps the price flat","summary":"If you capped how many Claude calls your agent could make to control spend, redo that budget. The same money now buys a meaningfully smarter model, so the coding assistant or research agent you throttled last quarter can run wider without a bigger invoice. Test it before you assume your old cost ceilings still hold.","analysis":"The headline number is the price that did not move. Anthropic shipped Opus 5 on July 24 at five dollars per million input tokens and twenty-five per million output, the same sticker as Opus 4.8, while claiming it more than doubles the older model's score on its internal Frontier-Bench and triples the next-best result on ARC-AGI 3, a reasoning test built to resist memorization. On coding it lands within half a percent of Fable 5, a rival frontier model, at half the cost per task. Capability is climbing a step function while the bill per token flattens, which quietly changes the math on every agent you were told to ration.","url":"https://www.anthropic.com/news/claude-opus-5","source":"Anthropic","date":"2026-07-25","tags":["models"]},{"id":"2026-07-25--amd-helios-anthropic-mi450","title":"AMD's Helios rack lands a 2-gigawatt Anthropic order to chase Nvidia","summary":"You will not buy a Helios rack, but you will feel the second supplier. If AMD becomes a real alternative to Nvidia, the GPU shortage that has been rationing model access and inflating cloud bills eases a little. For an engineer waiting on capacity to deploy, more competition upstream is the first thing that ever helps.","analysis":"The customer list is the story. AMD used its Advancing AI event on July 23 to unveil Helios, a rack-scale system Lisa Su called the industry's highest-performance AI rack, and named Microsoft, OpenAI, Meta, Oracle and Anthropic as buyers. Anthropic's slice alone runs up to 2 gigawatts of Instinct MI450 GPUs, enough electricity to light a small city. Helios ships later this year and claims to beat Nvidia's Vera Rubin systems on several metrics AMD did not fully detail. Nvidia has owned the training rack for years, so the buyers hedging their supply is what actually shifts here.","url":"https://techcrunch.com/2026/07/23/amd-takes-on-nvidia-with-its-helios-ai-rack-scale-system/","source":"TechCrunch","date":"2026-07-25","tags":["industry"]},{"id":"2026-07-25--prentis-lab-office-agents","title":"Prentis: Hoffman and Pincus back a lab training AI on office busywork","summary":"If your job is mostly moving data between systems, forms, spreadsheets, internal portals, this is the category aimed squarely at it. The people who stay valuable will own the judgment calls the software cannot make, not the clicks. Worth asking which parts of your week are pure navigation.","analysis":"A model that watches how a billing clerk clicks through her afternoon is the bet here. Prentis, co-founded by Ritankar Das with Reid Hoffman and Mark Pincus, is raising $100 million at a $1 billion valuation to build agents that learn routine office workflows by imitation. Its Hive-32B model already claims to beat OpenAI's GPT-5.4 and Anthropic's Claude Opus 4.6 on computer-use tests, where a model drives real software, at roughly a tenth of the API cost, and the eight-month-old company says it has signed contracts worth up to $50 million across healthcare, manufacturing and apparel. The target sits below the chat window, in the desk work itself.","url":"https://techcrunch.com/2026/07/24/prentis-new-ai-lab-co-founded-by-reid-hoffman-mark-pincus-in-talks-to-raise-100m/","source":"TechCrunch","date":"2026-07-25","tags":["agents","industry"]},{"id":"2026-07-25--aegisai-spear-phishing-defense","title":"AegisAI raises $36M as AI-written phishing doubles its hit rate","summary":"That polished email from your CEO asking for a quick favor is now cheap to fake at scale. For anyone in finance or IT, treat the personalized, well-written message as the one to verify out of band, by a call or a second channel. The grammar mistakes that used to give scams away are gone.","analysis":"The attackers automated first, so the defense is racing to catch up. AegisAI, founded by two former Google security engineers who worked on Safe Browsing and reCAPTCHA, raised a $36 million round led by Battery Ventures to build agents that scan email for the tells of AI-generated spear phishing, the targeted scam messages crafted for one specific recipient. Co-founder Cy Khormaee says AI-powered attacks now slip past existing filters more than half the time, close to twice as effective as the older kind. Customers already include LangChain and the fintech Mesh. The same tools writing the lures are being turned to catch them.","url":"https://techcrunch.com/2026/07/23/aegisai-founded-by-former-google-security-execs-lands-36m-to-stop-ai-driven-spear-phishing/","source":"TechCrunch","date":"2026-07-25","tags":["agents"]},{"id":"2026-07-25--runway-media-router-launch","title":"Runway launches a Media Router to pick the best image and video model","summary":"Building anything that generates images or audio? Stop hard-coding one vendor and route by the job instead. Locking to a single model in a field this fast is how you end up overpaying or stuck on a version everyone else has moved past. Abstraction over any one provider is quietly becoming table stakes.","analysis":"Choosing a generative model has become a full-time job, so Runway is automating the choice. On July 23 the company launched Media Router through its developer platform, a tool that sends each image, video or audio request to whichever model, its own or a third party's, best fits a developer's stated priority of quality, speed or cost. Runway leans on its in-house creative team's taste to score things like how a video model handles motion or lip-sync. Six launch customers including Adobe, Cloudflare, ElevenLabs and Expedia are already wiring it in to generate media inside their own products. When the models multiply, the router becomes the product.","url":"https://techcrunch.com/2026/07/23/runway-bets-on-ai-model-routing-as-generative-media-gets-crowded/","source":"TechCrunch","date":"2026-07-25","tags":["agents"]},{"id":"2026-07-25--etched-inference-chip-valuation","title":"Etched hits $10.3B on a bet that inference needs its own chip","summary":"For anyone paying an inference bill, more specialized chips is good news on price over the next two years. It also signals where the money thinks the market is heading: training was the gold rush, running the models is the durable business. If you deploy AI features, watch for cheaper inference-only options to reach your cloud console.","analysis":"Doubling your valuation in seven months on a chip that has not shipped in volume takes some nerve. Etched closed a $300 million round on July 23 at a $10.3 billion mark, up from $5 billion in December, with Sequoia leading and Peter Thiel, Andrej Karpathy and SK Hynix in the mix. The company builds silicon tuned only for inference, the work of running a trained model for users, splitting the job into a low-voltage prefill chip and a shared memory fabric that lets many chips pool memory, and it says it already holds $1 billion in booked orders with TSMC making the parts. The wager is that general-purpose GPUs waste money once a model is trained.","url":"https://techcrunch.com/2026/07/23/ai-chip-startup-etched-defies-skeptics-hits-10-3b-valuation-from-big-name-investors/","source":"TechCrunch","date":"2026-07-25","tags":["industry"]},{"id":"2026-07-25--kalanick-atoms-robotics-raise","title":"Travis Kalanick's Atoms raises $1.7B to push AI into mining and robots","summary":"Physical-world automation is where a lot of the next hiring wave lands, so a robotics technician or industrial engineer may see more demand, not less. This money chases work software has barely touched: mines, warehouses, factory floors. If your job is bits, the quiet story is how fast the frontier is moving to atoms.","analysis":"Uber's founder wants to automate the physical world, and a16z just handed him $1.7 billion to try. Travis Kalanick's holding company Atoms, built on the bones of his CloudKitchens venture, raised the round on July 22 with Andreessen Horowitz leading, Ben Horowitz taking a board seat, and Uber itself among the backers. Atoms bought Pronto, a heavy-industry automation firm, and plans to move into mining and robotics, with Kalanick describing manufacturing as compute and real estate as storage. Most of the cash is earmarked for hiring. The bet is that AI's next decade runs on trucks and drills more than chatbots.","url":"https://techcrunch.com/2026/07/22/travis-kalanicks-robotics-company-raises-1-7b-led-by-a16z/","source":"TechCrunch","date":"2026-07-25","tags":["industry"]},{"id":"2026-07-25--monday-com-ai-layoffs","title":"Monday.com cuts 630 jobs, a fifth of staff, to reshape around AI","summary":"When a growing company cuts jobs, the reorg tells you more than the severance does. If you work in sales or support at a software firm, watch for the phrase leaner operating model, which increasingly means agents are being slotted into the workflow you own. Ask where you sit in that redesign before it gets announced.","analysis":"A profitable software company shedding a fifth of its people is the shape 2026 layoffs keep taking. Monday.com cut roughly 630 workers on July 22, about 20 percent of headcount, and booked $45 million to $55 million in restructuring charges, saying it wants a leaner operation built around its AI work platform of agents and no-code automation. The firm is not in financial trouble. It rebuilt the product around AI agents working next to employees, then decided it needed fewer employees to sell that story. Across tech, 78 percent of this year's cuts cited an AI refocus. The rebrand and the redundancy arrive together.","url":"https://techcrunch.com/2026/07/22/monday-com-lays-off-hundreds-to-focuses-on-ai/","source":"TechCrunch","date":"2026-07-25","tags":["industry"]},{"id":"2026-07-25--eu-ai-act-high-risk-deadline","title":"EU AI Act's high-risk rules take effect August 2, and the clock is real","summary":"If your product touches EU users, this belongs on a roadmap now, not next quarter. A developer at a mid-size firm should know whether any feature counts as high-risk, hiring, credit scoring, biometrics, and what documentation that triggers. The rules phase in over years, but pleading ignorance stops working in nine days.","analysis":"The grace period ends in nine days. On August 2 the European Union's AI Act, its risk-based law for artificial intelligence, switches on the obligations for high-risk systems and general-purpose models, covering conformity checks, transparency notices and record-keeping. The Commission paired the date with a July action plan on AI and cybersecurity and a push to build EU capacity to test frontier models before they reach the market. Full enforcement against some high-risk systems embedded in products stretches to 2027, so the teeth arrive in stages. Any team shipping AI into Europe now works against a compliance calendar with a real deadline.","url":"https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai","source":"European Commission","date":"2026-07-25","tags":["industry"]}]}