https://aiinsiders.net/edition/2026-09-25 AI Insiders en 2026-09-24T19:35:30.051Z The Day an AI Agent Broke Into a Government Portal, and Two Labs Graded Their Own Work https://aiinsiders.net/article/ai-benchmark-forecasters-keep-guessing-years-too-slow-study AI Insiders en 2026-09-24T19:16:31.463Z AI benchmark forecasters keep guessing years too slow, study finds https://aiinsiders.net/article/alibabas-qwen-launches-three-ai-agents-to-run-your-phone AI Insiders en 2026-09-24T19:16:31.463Z Alibaba's Qwen Launches Three AI Agents to Run Your Phone https://aiinsiders.net/article/an-openai-agent-broke-into-an-australian-health-data-portal AI Insiders en 2026-09-24T19:16:31.463Z An OpenAI Agent Broke Into an Australian Health Data Portal https://aiinsiders.net/article/anthropic-says-claude-found-a-new-crispr-like-dna-system AI Insiders en 2026-09-24T19:16:31.463Z Anthropic says Claude found a new CRISPR-like DNA system https://aiinsiders.net/article/comfy-launches-one-api-for-calling-every-frontier-media AI Insiders en 2026-09-24T19:16:31.463Z Comfy launches one API for calling every frontier media model https://aiinsiders.net/article/fireworks-ships-a-trimmed-kimi-k3-that-burns-40percent AI Insiders en 2026-09-24T19:16:31.463Z Fireworks Ships a Trimmed Kimi K3 That Burns 40% Fewer Tokens https://aiinsiders.net/article/four-ai-models-broke-perplexitys-sandbox-network-rules-not AI Insiders en 2026-09-24T19:16:31.463Z Four AI models broke Perplexity's sandbox network rules, not its VM wall https://aiinsiders.net/article/gemini-can-now-reach-into-linear-adobe-webflow-and-peloton AI Insiders en 2026-09-24T19:16:31.463Z Gemini can now reach into Linear, Adobe, Webflow and Peloton https://aiinsiders.net/article/google-details-a-memory-system-it-says-even-google-cant-read AI Insiders en 2026-09-24T19:16:31.463Z Google details a memory system it says even Google can't read https://aiinsiders.net/article/googles-new-gemini-voice-models-can-clone-a-voice-from-30 AI Insiders en 2026-09-24T19:16:31.463Z Google's new Gemini voice models can clone a voice from 30 seconds https://aiinsiders.net/article/openai-built-a-mental-health-benchmark-then-graded-it-with AI Insiders en 2026-09-24T19:16:31.463Z OpenAI built a mental health benchmark, then graded it with its own model https://aiinsiders.net/article/perry-dong-says-robots-need-their-own-fine-tuning-playbook AI Insiders en 2026-09-24T19:16:31.463Z Perry Dong Says Robots Need Their Own Fine-Tuning Playbook https://aiinsiders.net/article/replits-ceo-argues-college-should-be-training-students-to AI Insiders en 2026-09-24T19:16:31.463Z Replit's CEO argues college should be training students to question everything https://aiinsiders.net/article/together-ai-ships-a-tiny-classifier-model-that-cost AI Insiders en 2026-09-24T19:16:31.463Z Together AI ships a tiny classifier model that cost $17 to train https://aiinsiders.net/edition/2026-09-24 AI Insiders en 2026-09-23T17:22:01.030Z The Day Two Rivals Both Cut Their Best Model's Price https://aiinsiders.net/article/a-memory-fix-lets-ai-training-reach-million-token-context AI Insiders en 2026-09-23T17:05:21.507Z A memory fix lets AI training reach million-token context https://aiinsiders.net/article/a-quieter-way-ai-firms-could-train-on-your-work-anyway AI Insiders en 2026-09-23T17:05:21.507Z A quieter way AI firms could train on your work anyway https://aiinsiders.net/article/a-startup-grew-ai-on-living-brain-cells-its-selling AI Insiders en 2026-09-23T17:05:21.507Z A startup grew AI on living brain cells. It's selling software, not neurons. https://aiinsiders.net/article/a-vcs-week-in-beijing-chinese-ai-labs-export-or-the AI Insiders en 2026-09-23T17:05:21.507Z A VC's week in Beijing: Chinese AI labs export or the grinding was for nothing https://aiinsiders.net/article/altman-and-amodei-set-to-brief-un-security-council-on-ai AI Insiders en 2026-09-23T17:05:21.507Z Altman and Amodei set to brief UN Security Council on AI risk https://aiinsiders.net/article/an-ai-model-just-cracked-a-decades-old-enigma-cipher-on-its AI Insiders en 2026-09-23T17:05:21.507Z An AI model just cracked a decades-old Enigma cipher on its own https://aiinsiders.net/article/anthropic-breaks-down-what-a-claude-code-task-actually AI Insiders en 2026-09-23T17:05:21.507Z Anthropic breaks down what a Claude Code task actually costs on Opus 5.5 https://aiinsiders.net/article/anthropic-ships-claude-opus-55-cheaper-and-safer-than-opus-5 AI Insiders en 2026-09-23T17:05:21.507Z Anthropic ships Claude Opus 5.5, cheaper and safer than Opus 5 https://aiinsiders.net/article/chinas-biggest-dram-maker-claims-it-has-matched-samsung-and AI Insiders en 2026-09-23T17:05:21.507Z China's biggest DRAM maker claims it has matched Samsung and Micron https://aiinsiders.net/article/ema-raises-dollar77m-to-replace-enterprise-software-with-ai AI Insiders en 2026-09-23T17:05:21.507Z Ema raises $77M to replace enterprise software with AI agent teams https://aiinsiders.net/article/google-researchers-built-a-fix-for-ai-agents-that-cheat AI Insiders en 2026-09-23T17:05:21.507Z Google researchers built a fix for AI agents that cheat their own tests https://aiinsiders.net/article/khosla-says-personal-ai-will-win-on-trust-not-features AI Insiders en 2026-09-23T17:05:21.507Z Khosla says personal AI will win on trust, not features https://aiinsiders.net/article/meta-concedes-muse-copied-openclaws-exact-agent-file-setup AI Insiders en 2026-09-23T17:05:21.507Z Meta concedes Muse copied OpenClaw's exact agent file setup https://aiinsiders.net/article/on-a-harder-coding-test-top-ai-models-solve-barely-one-in AI Insiders en 2026-09-23T17:05:21.507Z On a harder coding test, top AI models solve barely one in four bugs https://aiinsiders.net/article/openai-adds-a-dashboard-so-developers-can-see-their-gpt-6 AI Insiders en 2026-09-23T17:05:21.507Z OpenAI adds a dashboard so developers can see their GPT-6 cache savings https://aiinsiders.net/article/openai-halves-api-prices-with-two-cheaper-gpt-6-models AI Insiders en 2026-09-23T17:05:21.507Z OpenAI halves API prices with two cheaper GPT-6 models https://aiinsiders.net/article/perplexity-trains-its-ai-agent-on-its-own-mistakes-not-just AI Insiders en 2026-09-23T17:05:21.507Z Perplexity trains its AI agent on its own mistakes, not just its wins https://aiinsiders.net/article/vllm-adds-a-translator-layer-so-one-ai-model-runs-on-any AI Insiders en 2026-09-23T17:05:21.507Z vLLM adds a translator layer so one AI model runs on any chip https://aiinsiders.net/article/youtube-lets-you-describe-the-feed-you-want-instead-of AI Insiders en 2026-09-23T17:05:21.507Z YouTube lets you describe the feed you want instead of watching your way there https://aiinsiders.net/edition/2026-09-23 AI Insiders en 2026-09-22T21:23:10.859Z When the Only Judge Is the Company That Built It https://aiinsiders.net/article/a-writer-says-ai-models-are-about-to-split-into-cheap-and AI Insiders en 2026-09-22T21:09:23.573Z A writer says AI models are about to split into cheap and pricey parts https://aiinsiders.net/article/adding-more-ai-agents-buys-speed-not-smarts-new-math-shows AI Insiders en 2026-09-22T21:09:23.573Z Adding more AI agents buys speed, not smarts, new math shows https://aiinsiders.net/article/aikido-shrinks-a-15tb-security-model-to-run-entirely-on AI Insiders en 2026-09-22T21:09:23.573Z Aikido shrinks a 1.5TB security model to run entirely on your own servers https://aiinsiders.net/article/alibaba-unveils-new-ai-chip-says-it-triples-predecessors AI Insiders en 2026-09-22T21:09:23.573Z Alibaba unveils new AI chip, says it triples predecessor's speed https://aiinsiders.net/article/astroforge-will-let-its-own-ai-model-fly-the-next-asteroid AI Insiders en 2026-09-22T21:09:23.573Z AstroForge will let its own AI model fly the next asteroid mission https://aiinsiders.net/article/aws-ships-an-agent-harness-that-runs-any-ai-model-you-pick AI Insiders en 2026-09-22T21:09:23.573Z AWS Ships an Agent Harness That Runs Any AI Model You Pick https://aiinsiders.net/article/chinas-open-ai-models-are-pulling-ahead-of-americas-open AI Insiders en 2026-09-22T21:09:23.573Z China's open AI models are pulling ahead of America's open models https://aiinsiders.net/article/claude-users-say-theyve-spotted-an-unreleased-anthropic AI Insiders en 2026-09-22T21:09:23.573Z Claude users say they've spotted an unreleased Anthropic model https://aiinsiders.net/article/cognition-brings-its-devin-coding-agents-cloud-sessions AI Insiders en 2026-09-22T21:09:23.573Z Cognition brings its Devin coding agent's cloud sessions into the terminal https://aiinsiders.net/article/developer-open-sources-small-ai-models-that-answer-yesno AI Insiders en 2026-09-22T21:09:23.573Z Developer open sources small AI models that answer yes/no questions locally https://aiinsiders.net/article/ex-anthropic-teams-startup-mirendil-in-talks-at-dollar5 AI Insiders en 2026-09-22T21:09:23.573Z Ex-Anthropic team's startup Mirendil in talks at $5 billion, sources say https://aiinsiders.net/article/mathematicians-push-back-on-openai-so-it-built-an-outside AI Insiders en 2026-09-22T21:09:23.573Z Mathematicians push back on OpenAI, so it built an outside advisory group https://aiinsiders.net/article/metas-muse-agent-app-beat-chatgpt-on-downloads-then-hit-an AI Insiders en 2026-09-22T21:09:23.573Z Meta's Muse agent app beat ChatGPT on downloads, then hit an Amazon wall https://aiinsiders.net/article/qwens-new-agent-benchmark-shows-top-models-still-fail-most AI Insiders en 2026-09-22T21:09:23.573Z Qwen's new agent benchmark shows top models still fail most rebuilds https://aiinsiders.net/article/stepfuns-step-5-matches-kimi-k3-at-far-lower-cost-testing AI Insiders en 2026-09-22T21:09:23.573Z StepFun's Step 5 Matches Kimi K3 at Far Lower Cost, Testing Finds https://aiinsiders.net/article/tomasz-tunguz-says-tiny-ai-models-are-quietly-replacing-old AI Insiders en 2026-09-22T21:09:23.573Z Tomasz Tunguz Says Tiny AI Models Are Quietly Replacing Old Code Rules https://aiinsiders.net/article/why-ai-labs-are-racing-into-ads-robots-and-cancer-research AI Insiders en 2026-09-22T21:09:23.573Z Why AI Labs Are Racing Into Ads, Robots and Cancer Research https://aiinsiders.net/article/xai-ships-grok-47-but-its-own-scorecard-shows-a-mixed-lead AI Insiders en 2026-09-22T21:09:23.573Z xAI ships Grok 4.7, but its own scorecard shows a mixed lead https://aiinsiders.net/article/xiaomi-open-sources-a-frontier-ai-model-and-undercuts-on AI Insiders en 2026-09-22T21:09:23.573Z Xiaomi open sources a frontier AI model, and undercuts on price https://aiinsiders.net/article/xiaomi-says-grading-code-quality-not-just-passfail-fixed AI Insiders en 2026-09-22T21:09:23.573Z Xiaomi says grading code quality, not just pass/fail, fixed its AI training