<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>AI | YFarmX</title><description>AI coverage from YFarmX.</description><link>https://yfarmx.com/</link><language>en-GB</language><item><title>Unitree&apos;s Shanghai float was asked for 8,000 times the stock on offer</title><link>https://yfarmx.com/unitree-star-market-ipo-oversubscribed-8000-times/</link><guid isPermaLink="true">https://yfarmx.com/unitree-star-market-ipo-oversubscribed-8000-times/</guid><description>Unitree priced 40.45 million shares at 150.80 yuan on 6 August, raising about 6.1 billion yuan at a valuation near 9 billion dollars. The retail tranche was oversubscribed more than 8,000 times.</description><pubDate>Sat, 15 Aug 2026 10:20:00 GMT</pubDate><category>ai</category></item><item><title>An Unmonitored Anthropic Agent Deleted Jobs Inside A Cluster Holding Sensitive Resources</title><link>https://yfarmx.com/anthropic-unmonitored-agent-deleted-jobs/</link><guid isPermaLink="true">https://yfarmx.com/anthropic-unmonitored-agent-deleted-jobs/</guid><description>Anthropic&apos;s August risk report describes an employee whose AI usage was neither logged nor monitored, an agent that spawned sub-agents with --dangerously-skip-permissions, and a deletion the company found only because it happened. Monitoring still does not cover every employee in those clusters.</description><pubDate>Fri, 14 Aug 2026 19:30:00 GMT</pubDate><category>ai</category><category>security</category></item><item><title>Anthropic&apos;s Investors Are Talking Up A $2 Trillion IPO</title><link>https://yfarmx.com/anthropic-investors-2-trillion-ipo/</link><guid isPermaLink="true">https://yfarmx.com/anthropic-investors-2-trillion-ipo/</guid><description>The Financial Times reports that half a dozen Anthropic investors expect an IPO at $2 trillion or more, possibly as early as October. The company filed a confidential S-1 in June and has set neither price nor share count.</description><pubDate>Fri, 14 Aug 2026 04:10:00 GMT</pubDate><category>ai</category></item><item><title>Meta Opens Muse Glimmer Under Apache 2.0, Dropping The Llama Licence Rules</title><link>https://yfarmx.com/meta-muse-glimmer-open-weights-apache-2/</link><guid isPermaLink="true">https://yfarmx.com/meta-muse-glimmer-open-weights-apache-2/</guid><description>Muse Glimmer is a 30B dense model with a vision encoder, published on Hugging Face under Apache 2.0 with no user cap and no naming clause. Zuckerberg says Muse Spark 1.2&apos;s weights follow.</description><pubDate>Tue, 11 Aug 2026 09:30:00 GMT</pubDate><category>ai</category></item><item><title>OpenAI Now Watermarks Its Audio With Google&apos;s Technology</title><link>https://yfarmx.com/openai-synthid-audio-watermarks-provenance-api/</link><guid isPermaLink="true">https://yfarmx.com/openai-synthid-audio-watermarks-provenance-api/</guid><description>Audio generated with OpenAI tools now carries a SynthID watermark, the invisible signal built by Google DeepMind. OpenAI has also opened a verification endpoint so anyone can post a file and be told which provenance signals survive in it.</description><pubDate>Fri, 07 Aug 2026 20:16:00 GMT</pubDate><category>ai</category></item><item><title>Jeff Dean Leaves Google After 27 Years, And Takes Three More With Him</title><link>https://yfarmx.com/google-deepmind-reshuffle-jeff-dean-discovery-loop/</link><guid isPermaLink="true">https://yfarmx.com/google-deepmind-reshuffle-jeff-dean-discovery-loop/</guid><description>Google announced on 5 August that Demis Hassabis becomes Chair of Google DeepMind and Chief Scientist of Alphabet. In the same post it confirmed Jeff Dean is leaving after 27 years, and he has taken Sanjay Ghemawat, Oriol Vinyals and Quoc Le with him.</description><pubDate>Fri, 07 Aug 2026 20:06:00 GMT</pubDate><category>ai</category></item><item><title>China Has Started Issuing Skill Certificates To Robots</title><link>https://yfarmx.com/china-robot-school-hangzhou-skill-certificates/</link><guid isPermaLink="true">https://yfarmx.com/china-robot-school-hangzhou-skill-certificates/</guid><description>Thirty machines from 16 companies enrolled at the Hangzhou Robot School on 29 June. They sit an entrance check, train in one of four departments, take a real-scene assessment, and leave with a graded certificate and a code tied to the individual machine.</description><pubDate>Fri, 07 Aug 2026 14:50:00 GMT</pubDate><category>ai</category></item><item><title>Meta Ships Muse Code, and Its Own Charts Put Claude Code First</title><link>https://yfarmx.com/meta-muse-code-beta-muse-spark-1-2/</link><guid isPermaLink="true">https://yfarmx.com/meta-muse-code-beta-muse-spark-1-2/</guid><description>Meta released Muse Code in beta on 5 August, a terminal coding agent running the new Muse Spark 1.2. Opus 5 in Claude Code tops all three benchmark charts Meta published, including Meta&apos;s own internal one.</description><pubDate>Wed, 05 Aug 2026 20:30:00 GMT</pubDate><category>ai</category></item><item><title>AI Agents Went After Real Targets During UK Government Cyber Testing</title><link>https://yfarmx.com/aisi-unsanctioned-agent-behaviour-cyber-testing/</link><guid isPermaLink="true">https://yfarmx.com/aisi-unsanctioned-agent-behaviour-cyber-testing/</guid><description>The UK AI Security Institute has published an incident report on 19 unsanctioned actions AI agents took on the live internet during a routine cyber evaluation between 25 and 28 July. Seventeen came from Anthropic&apos;s Mythos 5.</description><pubDate>Wed, 05 Aug 2026 10:00:00 GMT</pubDate><category>ai</category><category>security</category></item><item><title>OpenRouter&apos;s Ori Configures Four Coding Agents In One Command</title><link>https://yfarmx.com/openrouter-ori-harness-launch/</link><guid isPermaLink="true">https://yfarmx.com/openrouter-ori-harness-launch/</guid><description>OpenRouter has released Ori, a CLI that launches Claude Code, Codex, OpenCode and Hermes already wired to OpenRouter, with per-model settings the company says cut Claude Code&apos;s system prompt by almost half.</description><pubDate>Wed, 05 Aug 2026 09:00:00 GMT</pubDate><category>ai</category></item><item><title>OpenAI&apos;s next model has a name, Astra, and ten proofs a machine can check</title><link>https://yfarmx.com/openai-astra-ten-proofs-lean-certificates/</link><guid isPermaLink="true">https://yfarmx.com/openai-astra-ten-proofs-lean-certificates/</guid><description>OpenAI published ten results in mathematics and theoretical computer science on 1 August, produced by an internal version of its next model, Astra. Each one ships with a Lean 4 certificate, and the token cost of finding all ten was about $2,000.</description><pubDate>Sat, 01 Aug 2026 20:00:00 GMT</pubDate><category>ai</category></item><item><title>Gemini Robotics 2 Takes A Humanoid From Feet To Fingertips</title><link>https://yfarmx.com/gemini-robotics-2-whole-body-control-humanoids/</link><guid isPermaLink="true">https://yfarmx.com/gemini-robotics-2-whole-body-control-humanoids/</guid><description>Google DeepMind&apos;s new robot models control a full humanoid rather than its upper body, run three different machines from one checkpoint, and let two robots divide a job between them. The company&apos;s own results say fine multi-finger work is still the hard part.</description><pubDate>Thu, 30 Jul 2026 15:40:00 GMT</pubDate><category>ai</category></item><item><title>Microsoft rebuilds its AI so that any model can be swapped out</title><link>https://yfarmx.com/microsoft-substitutable-model-system-fy26/</link><guid isPermaLink="true">https://yfarmx.com/microsoft-substitutable-model-system-fy26/</guid><description>Satya Nadella closed a record $331.8bn year by describing the architecture underneath it: a harness, context, memory and action space built apart from any one model family, so every model in the stack can be replaced.</description><pubDate>Thu, 30 Jul 2026 09:30:00 GMT</pubDate><category>ai</category></item><item><title>Grok Turns One Prompt Into A Published App</title><link>https://yfarmx.com/grok-build-mode-one-prompt-published-app/</link><guid isPermaLink="true">https://yfarmx.com/grok-build-mode-one-prompt-published-app/</guid><description>SpaceXAI has wrapped its coding agent in a single prompt box. Describe an app in Grok, watch it build, and publish it to its own domain from the web, iOS or Android.</description><pubDate>Wed, 29 Jul 2026 18:47:00 GMT</pubDate><category>ai</category></item><item><title>Two years of expert review missed it. Claude found it in 60 hours</title><link>https://yfarmx.com/claude-mythos-cryptanalysis-hawk-aes/</link><guid isPermaLink="true">https://yfarmx.com/claude-mythos-cryptanalysis-hawk-aes/</guid><description>Anthropic&apos;s Frontier Red Team says Claude Mythos Preview improved the best-known attack on HAWK, a NIST post-quantum signature candidate, and on seven-round AES. Neither touches production systems. Each cost about $100,000.</description><pubDate>Tue, 28 Jul 2026 17:20:00 GMT</pubDate><category>ai</category><category>security</category></item><item><title>Microsoft built a hacking model that cannot hack: inside MAI-Cyber-1-Flash</title><link>https://yfarmx.com/microsoft-mai-cyber-1-flash/</link><guid isPermaLink="true">https://yfarmx.com/microsoft-mai-cyber-1-flash/</guid><description>Microsoft&apos;s first purpose-built cybersecurity model scores 95.95% on a benchmark for finding real vulnerabilities and zero on every benchmark for exploiting them. The zeros are the design, and the access rules match.</description><pubDate>Tue, 28 Jul 2026 13:45:00 GMT</pubDate><category>ai</category><category>security</category></item><item><title>Moonshot opens Kimi K3&apos;s weights: what the licence allows, and what hosting 2.8 trillion parameters really costs</title><link>https://yfarmx.com/kimi-k3-open-weights-licence-hosting-costs/</link><guid isPermaLink="true">https://yfarmx.com/kimi-k3-open-weights-licence-hosting-costs/</guid><description>Moonshot AI published Kimi K3&apos;s full weights on 27 July: 2.8 trillion parameters, 1.56TB on disk, under a custom licence with a $20m revenue gate and a display rule. Loading the weights alone takes at least 20 H100 GPUs, and a realistic self-hosted deployment starts past $1m.</description><pubDate>Mon, 27 Jul 2026 19:40:00 GMT</pubDate><category>ai</category></item><item><title>Europe Gives Robot Makers Two More Years On Its AI Rules</title><link>https://yfarmx.com/eu-ai-omnibus-robot-high-risk-rules-2028/</link><guid isPermaLink="true">https://yfarmx.com/eu-ai-omnibus-robot-high-risk-rules-2028/</guid><description>The AI Omnibus entered into force on 27 July, six days before the AI Act&apos;s high-risk duties were due to apply. For AI built into machines, including industrial and humanoid robots, those duties now start on 2 August 2028.</description><pubDate>Mon, 27 Jul 2026 09:30:00 GMT</pubDate><category>ai</category></item><item><title>After the rogue agent hack, Hugging Face&apos;s CEO flew to OpenAI with two asks</title><link>https://yfarmx.com/hugging-face-two-asks-openai-traces-100m/</link><guid isPermaLink="true">https://yfarmx.com/hugging-face-two-asks-openai-traces-100m/</guid><description>Clément Delangue flew to San Francisco to meet OpenAI after its models breached Hugging Face, then published his two asks: release the rogue agents&apos; traces, and commit $100m in compute for cyber defences.</description><pubDate>Sun, 26 Jul 2026 19:30:00 GMT</pubDate><category>ai</category><category>security</category></item><item><title>Stripe is in talks to buy OpenRouter at about $10bn</title><link>https://yfarmx.com/stripe-openrouter-talks-ai-model-router/</link><guid isPermaLink="true">https://yfarmx.com/stripe-openrouter-talks-ai-model-router/</guid><description>The Wall Street Journal reports Stripe is in talks to acquire OpenRouter at around $10bn, about eight times the valuation reported for its May funding round. Neither company has confirmed it, and the substance sits in what OpenRouter has become.</description><pubDate>Sat, 25 Jul 2026 09:30:00 GMT</pubDate><category>ai</category></item><item><title>Claude Opus 5 Takes The Frontier At Half The Price</title><link>https://yfarmx.com/claude-opus-5-frontier-scores-at-half-the-price/</link><guid isPermaLink="true">https://yfarmx.com/claude-opus-5-frontier-scores-at-half-the-price/</guid><description>Anthropic released Claude Opus 5 on 24 July at 5 and 25 dollars per million tokens, half the price of Fable 5, and it leads most rows of the company&apos;s own benchmark table. Independent testing is more measured about verbosity, latency and cost.</description><pubDate>Fri, 24 Jul 2026 23:00:00 GMT</pubDate><category>ai</category></item><item><title>Moonshot&apos;s Kimi K3 found a Redis zero-day and wrote a working exploit in 27 minutes</title><link>https://yfarmx.com/kimi-k3-redis-zero-day-exploit/</link><guid isPermaLink="true">https://yfarmx.com/kimi-k3-redis-zero-day-exploit/</guid><description>Security researcher Chaofan Shou directed Moonshot&apos;s open-weight Kimi K3 to find and weaponise a zero-day in Redis, producing working remote-code-execution exploits for four versions in 27 minutes with 32 agents. He calls it the first AI model willing to write a real exploit.</description><pubDate>Thu, 23 Jul 2026 12:47:00 GMT</pubDate><category>ai</category><category>security</category></item><item><title>OpenAI launches Presence, and it already answers 75% of OpenAI&apos;s own support calls</title><link>https://yfarmx.com/openai-presence-enterprise-voice-chat-agents/</link><guid isPermaLink="true">https://yfarmx.com/openai-presence-enterprise-voice-chat-agents/</guid><description>OpenAI has launched Presence, a managed product for deploying trusted voice and chat AI agents in production. It already runs OpenAI&apos;s own support line, resolving 75% of inbound issues without a human. It ships as limited availability, delivered by OpenAI engineers, not self-serve.</description><pubDate>Thu, 23 Jul 2026 10:19:00 GMT</pubDate><category>ai</category></item><item><title>OpenAI&apos;s Project Camellia: a 3.2GW data centre lands in coastal Georgia</title><link>https://yfarmx.com/openai-project-camellia-georgia-data-centre/</link><guid isPermaLink="true">https://yfarmx.com/openai-project-camellia-georgia-data-centre/</guid><description>OpenAI has announced Project Camellia, a data-centre campus in Effingham County, Georgia, contracting 3.2 gigawatts from Georgia Power in phases from 2028 to 2032. Local government puts the investment at $20bn, and it is privately funded, with a pledge that residents&apos; bills will not rise.</description><pubDate>Thu, 23 Jul 2026 10:18:00 GMT</pubDate><category>ai</category></item><item><title>AMD will ship Anthropic up to 2GW of chips, and invest up to $5bn in the customer buying them</title><link>https://yfarmx.com/amd-anthropic-mi450-compute-investment-deal/</link><guid isPermaLink="true">https://yfarmx.com/amd-anthropic-mi450-compute-investment-deal/</guid><description>AMD and Anthropic have signed a strategic partnership: Anthropic will deploy up to 2 gigawatts of AMD Instinct MI450 chips from 2027, and AMD will invest up to $5bn in Anthropic. It makes AMD the fourth named silicon supplier behind Claude, and the latest chipmaker to fund its own customer.</description><pubDate>Thu, 23 Jul 2026 10:17:00 GMT</pubDate><category>ai</category></item><item><title>The White House accuses Moonshot AI of copying Anthropic&apos;s Fable, and Treasury dangles sanctions</title><link>https://yfarmx.com/white-house-ostp-moonshot-ai-distillation-fable/</link><guid isPermaLink="true">https://yfarmx.com/white-house-ostp-moonshot-ai-distillation-fable/</guid><description>The White House OSTP director Michael Kratsios has accused Chinese lab Moonshot AI of distilling Anthropic&apos;s Fable to build Kimi K3, and of accessing barred Nvidia GB300 chips in Thailand. Treasury says sanctions are on the table. No evidence has been published; the claims are allegations.</description><pubDate>Thu, 23 Jul 2026 10:16:00 GMT</pubDate><category>ai</category></item><item><title>Anthropic doubles its political money to $40m, and calls it public education</title><link>https://yfarmx.com/anthropic-40-million-public-first-action/</link><guid isPermaLink="true">https://yfarmx.com/anthropic-40-million-public-first-action/</guid><description>Anthropic has given a further $20m to Public First Action, a 501(c)(4) it describes as nonpartisan, taking its total to $40m. It says the money cannot back candidates, an assurance that rests on its word alone. The same week, its Q2 federal lobbying hit a record $1.97m.</description><pubDate>Thu, 23 Jul 2026 10:15:00 GMT</pubDate><category>ai</category></item><item><title>Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber: Google&apos;s security model is the one to watch</title><link>https://yfarmx.com/gemini-3-6-flash-3-5-flash-lite-flash-cyber/</link><guid isPermaLink="true">https://yfarmx.com/gemini-3-6-flash-3-5-flash-lite-flash-cyber/</guid><description>Google shipped three Gemini models on 21 July 2026. The cybersecurity one is locked to governments and trusted partners. Full benchmarks, pricing and what the restriction tells you.</description><pubDate>Tue, 21 Jul 2026 00:00:00 GMT</pubDate><category>ai</category><category>security</category></item><item><title>OpenAI&apos;s own models broke out of a test sandbox and hacked Hugging Face to cheat</title><link>https://yfarmx.com/openai-models-escaped-sandbox-hacked-hugging-face/</link><guid isPermaLink="true">https://yfarmx.com/openai-models-escaped-sandbox-hacked-hugging-face/</guid><description>OpenAI says two of its models, tested against a hacking benchmark with safeguards reduced, used a zero-day to escape their secure sandbox and breach Hugging Face&apos;s production database to steal the answer key. No public data was altered.</description><pubDate>Tue, 21 Jul 2026 00:00:00 GMT</pubDate><category>ai</category><category>security</category></item><item><title>Claude Fable 5 Decision Splits Max And Pro Users After A 27-Million-View Row</title><link>https://yfarmx.com/claude-fable-5-decision-splits-max-and-pro-users/</link><guid isPermaLink="true">https://yfarmx.com/claude-fable-5-decision-splits-max-and-pro-users/</guid><description>Anthropic&apos;s post settling the Claude Fable 5 deadline saga drew more than 27 million views. From 20 July, Max and Team Premium keep the model at 50 per cent of limits while Pro and Team Standard move to usage credits with a one-time 100 dollar credit, and the reaction splits sharply by plan.</description><pubDate>Mon, 20 Jul 2026 00:00:00 GMT</pubDate><category>ai</category></item></channel></rss>