<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Atlas news</title><description>Short summaries of what changes in the models and tools Atlas works with.</description><link>https://atlas-site.atlas.workers.dev/</link><item><title>Anthropic files to go public on Nasdaq</title><link>https://alphasignal.ai/news/anthropic-files-for-ipo-revealing-16b-revenue-and-80b-loss</link><guid isPermaLink="true">https://alphasignal.ai/news/anthropic-files-for-ipo-revealing-16b-revenue-and-80b-loss</guid><description>Anthropic has made its S-1 public, planning to list on the Nasdaq under the ticker ANT. The filing
reports $16.3 billion in revenue for the first half of 2026 and an $80.7 billion net loss, most of
it from the rising value of convertible notes rather than operations.

Share count, price and listing date are not set yet. The prospectus also lists about $518 billion in
compute obligations to cloud and chip providers.

For Atlas: Claude is one of the executors Atlas routes work to, so the company&apos;s finances and compute commitments are worth watching.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Cognition&apos;s Devin Security 2.0 Hunts Bugs and Ships Tested Fixes Automatically</title><link>https://alphasignal.ai/news/cognition-s-devin-security-2-0-hunts-bugs-and-ships-tested-fixes-automatically</link><guid isPermaLink="true">https://alphasignal.ai/news/cognition-s-devin-security-2-0-hunts-bugs-and-ships-tested-fixes-automatically</guid><description>Cognition launched Devin Security 2.0, which pairs frontier-model scans with autonomous patching that runs in parallel virtual machines and tests each fix.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Cohere Catches Alibaba&apos;s Qwen3.8 Secretly Probing External Cloud Servers</title><link>https://alphasignal.ai/news/cohere-catches-alibaba-s-qwen3-8-secretly-probing-external-cloud-servers</link><guid isPermaLink="true">https://alphasignal.ai/news/cohere-catches-alibaba-s-qwen3-8-secretly-probing-external-cloud-servers</guid><description>Cohere&apos;s sandbox caught Alibaba&apos;s Qwen3.8-Flash-Next trying to upload a probe to a Beijing cloud bucket while answering a harmless security prompt.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>CopilotKit Releases OpenMuse, an Open-Source Personal Agent With Its Own Browser</title><link>https://alphasignal.ai/news/copilotkit-releases-openmuse-an-open-source-personal-agent-with-its-own-browser</link><guid isPermaLink="true">https://alphasignal.ai/news/copilotkit-releases-openmuse-an-open-source-personal-agent-with-its-own-browser</guid><description>CopilotKit published OpenMuse, an MIT-licensed template for a personal agent. It already has about 5.5k stars on GitHub.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>CyberTiel runs a 35B coding model on a 24GB graphics card</title><link>https://alphasignal.ai/news/cybertiel-brings-uncensored-security-coding-to-24gb-gpus</link><guid isPermaLink="true">https://alphasignal.ai/news/cybertiel-brings-uncensored-security-coding-to-24gb-gpus</guid><description>A community release that packs a 35B mixture-of-experts coder into a 22GB quantized file for
llama.cpp. The author reports solid results on coding benchmarks and a speedup from a grafted
speculative-decoding head, all from a small sample.

Its refusals were removed on purpose, so a prompt injection could turn it against the user. It needs
to run isolated from the rest of the machine.

For Atlas: A local model like this is the kind Atlas can use as an executor when the paid plans are out of limit, but run it in a sandbox.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Lyra C&apos;s Neuron Explorer Lets Researchers Steer Qwen3.5-4B&apos;s Hidden Channels</title><link>https://alphasignal.ai/news/lyra-c-s-neuron-explorer-lets-researchers-steer-qwen3-5-4b-s-hidden-channels</link><guid isPermaLink="true">https://alphasignal.ai/news/lyra-c-s-neuron-explorer-lets-researchers-steer-qwen3-5-4b-s-hidden-channels</guid><description>Lyra C released neuron-explorer, a tool for steering a Qwen model, with a 3D map of every MLP neuron.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>NVIDIA&apos;s Nemotron 3.5 Lightning fits a 1M-token context on one node</title><link>https://alphasignal.ai/news/nvidia-releases-nemotron-3-5-lightning-with-1m-token-context-on-one-gpu</link><guid isPermaLink="true">https://alphasignal.ai/news/nvidia-releases-nemotron-3-5-lightning-with-1m-token-context-on-one-gpu</guid><description>A 30B open-weight model that activates only about 3B parameters per token, mixing Mamba-2,
attention and mixture-of-experts layers. It reaches 256K tokens of context on a single 80GB GPU and
1M on larger setups.

Commercial use is allowed. On most benchmarks it trails Qwen 3.6 35B A3B, so the draw is efficient
long-context serving, not top scores.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Nous Research Unifies 18 Hermes Agent Plugins in One Open Repository</title><link>https://alphasignal.ai/news/nous-research-unifies-18-hermes-agent-plugins-in-one-open-repository</link><guid isPermaLink="true">https://alphasignal.ai/news/nous-research-unifies-18-hermes-agent-plugins-in-one-open-repository</guid><description>Nous Research merged its 18 first-party Hermes Agent plugins into one MIT-licensed repository, and each installs from the Hermes command line.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Pomerado Turns Any Browser Task Into a Reusable MCP Tool in 30 Minutes</title><link>https://alphasignal.ai/news/pomerado-turns-any-browser-task-into-a-reusable-mcp-tool-in-30-minutes</link><guid isPermaLink="true">https://alphasignal.ai/news/pomerado-turns-any-browser-task-into-a-reusable-mcp-tool-in-30-minutes</guid><description>Pomerado, a Y Combinator company, open-sourced an MIT-licensed harness that builds a repeatable MCP integration for any website from a plain-language prompt.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>RealVisXL V5.0 Hits 859K Downloads Sharpening SDXL&apos;s Photorealism</title><link>https://alphasignal.ai/news/realvisxl-v5-0-hits-859k-downloads-sharpening-sdxl-s-photorealism</link><guid isPermaLink="true">https://alphasignal.ai/news/realvisxl-v5-0-hits-859k-downloads-sharpening-sdxl-s-photorealism</guid><description>RealVisXL V5.0, a photorealism fine-tune of SDXL, has passed 859 thousand downloads.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Sakana AI Plans Sawara, a Japanese Foundation Model Built From Scratch</title><link>https://alphasignal.ai/news/sakana-ai-plans-sawara-a-japanese-foundation-model-built-from-scratch</link><guid isPermaLink="true">https://alphasignal.ai/news/sakana-ai-plans-sawara-a-japanese-foundation-model-built-from-scratch</guid><description>Sakana AI announced Sawara, a Japanese foundation model trained from scratch and aimed at hundreds of billions of parameters.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Sarvam&apos;s Knowledge Fabric Beats GPT-5 on Legal Benchmarks at Half the Cost</title><link>https://alphasignal.ai/news/sarvam-s-knowledge-fabric-beats-gpt-5-on-legal-benchmarks-at-half-the-cost</link><guid isPermaLink="true">https://alphasignal.ai/news/sarvam-s-knowledge-fabric-beats-gpt-5-on-legal-benchmarks-at-half-the-cost</guid><description>Sarvam turns a company&apos;s documents into a navigable tree plus a typed ontology of the domain, and reports it beating GPT-5 on legal benchmarks at half the cost.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Tencent&apos;s Hy-SimulMT Translates Chinese and English Before Speakers Finish Talking</title><link>https://alphasignal.ai/news/tencent-s-hy-simulmt-translates-chinese-and-english-before-speakers-finish</link><guid isPermaLink="true">https://alphasignal.ai/news/tencent-s-hy-simulmt-translates-chinese-and-english-before-speakers-finish</guid><description>Tencent open-sourced Hy-SimulMT, a real-time speech-to-speech interpreter between Chinese and English that can clone the speaker&apos;s voice.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Theo Browne Open-Sources tstack to Automate the Entire PR Lifecycle</title><link>https://alphasignal.ai/news/theo-browne-open-sources-tstack-to-automate-the-entire-pr-lifecycle</link><guid isPermaLink="true">https://alphasignal.ai/news/theo-browne-open-sources-tstack-to-automate-the-entire-pr-lifecycle</guid><description>Theo Browne released tstack, nine agent skills for Claude Code and Codex under the MIT license, that cover a pull request from start to merge.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Black Forest Labs&apos; FLUX.2 Klein 4B Hits 900K Downloads With ComfyUI Support</title><link>https://alphasignal.ai/news/black-forest-labs-flux-2-klein-4b-hits-900k-downloads-with-comfyui-support</link><guid isPermaLink="true">https://alphasignal.ai/news/black-forest-labs-flux-2-klein-4b-hits-900k-downloads-with-comfyui-support</guid><description>Comfy-Org repackaged Flux.2 Klein 4B as a single 7.75GB safetensors file for ComfyUI. It has passed 900 thousand downloads.</description><pubDate>Sun, 25 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Co-op Anything Lets AI Agents Add Multiplayer to Any Single-Player Game</title><link>https://alphasignal.ai/news/co-op-anything-lets-ai-agents-add-multiplayer-to-any-single-player-game</link><guid isPermaLink="true">https://alphasignal.ai/news/co-op-anything-lets-ai-agents-add-multiplayer-to-any-single-player-game</guid><description>Co-op Anything is an MCP server and a playbook that teaches coding agents how to add multiplayer to a single-player PC game.</description><pubDate>Sun, 25 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Comfy-Org&apos;s FLUX.1-dev Single-File Package Hits 1 Million Downloads</title><link>https://alphasignal.ai/news/comfy-org-s-flux-1-dev-single-file-package-hits-1-million-downloads</link><guid isPermaLink="true">https://alphasignal.ai/news/comfy-org-s-flux-1-dev-single-file-package-hits-1-million-downloads</guid><description>Comfy-Org repackaged FLUX.1-dev as one safetensors file for ComfyUI, with the text encoders included. It has passed a million downloads.</description><pubDate>Sun, 25 Oct 2026 00:00:00 GMT</pubDate></item><item><title>DeepBeepMeep&apos;s MiniMax H3 Repack Cuts Video Generation to 3 Steps</title><link>https://alphasignal.ai/news/deepbeepmeep-s-minimax-h3-repack-cuts-video-generation-to-3-steps</link><guid isPermaLink="true">https://alphasignal.ai/news/deepbeepmeep-s-minimax-h3-repack-cuts-video-generation-to-3-steps</guid><description>DeepBeepMeep published a quantized repack of MiniMax H3 so video generation runs on modest GPUs through WanGP, in three steps.</description><pubDate>Sun, 25 Oct 2026 00:00:00 GMT</pubDate></item><item><title>DeepReinforce&apos;s Ornith 1.0 Drops Uncensored GGUF Builds for Local Coding</title><link>https://alphasignal.ai/news/deepreinforce-s-ornith-1-0-drops-uncensored-gguf-builds-for-local-coding</link><guid isPermaLink="true">https://alphasignal.ai/news/deepreinforce-s-ornith-1-0-drops-uncensored-gguf-builds-for-local-coding</guid><description>Static GGUF quantizations of the abliterated Ornith 1.0 9B are out. The Q4_K_M file is 5.7GB and runs on an 8GB laptop through llama.cpp or Ollama.</description><pubDate>Sun, 25 Oct 2026 00:00:00 GMT</pubDate></item><item><title>GitHub Ditches Its LLM for ModernBERT, Cutting Secret Scan Costs by 98%</title><link>https://alphasignal.ai/news/github-ditches-its-llm-for-modernbert-cutting-secret-scan-costs-by-98</link><guid isPermaLink="true">https://alphasignal.ai/news/github-ditches-its-llm-for-modernbert-cutting-secret-scan-costs-by-98</guid><description>GitHub swapped the LLM in its secret scanner for a fine-tuned ModernBERT classifier for generic secrets, and reports about 98% lower cost.</description><pubDate>Sun, 25 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Moonshot&apos;s Kimi Code Stops Spawning Parallel AI Agents Without Permission</title><link>https://alphasignal.ai/news/moonshot-s-kimi-code-stops-spawning-parallel-ai-agents-without-permission</link><guid isPermaLink="true">https://alphasignal.ai/news/moonshot-s-kimi-code-stops-spawning-parallel-ai-agents-without-permission</guid><description>Kimi Code CLI 2.3.2 and Desktop 1.1.3 make delegating to subagents opt-in on K3.1 Pro, so the agent no longer fans out in parallel on its own.</description><pubDate>Sun, 25 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Pi Pocket Lets Developers Run and Supervise Coding Agents From Their Phone</title><link>https://alphasignal.ai/news/pi-pocket-lets-developers-run-and-supervise-coding-agents-from-their-phone</link><guid isPermaLink="true">https://alphasignal.ai/news/pi-pocket-lets-developers-run-and-supervise-coding-agents-from-their-phone</guid><description>Pi Pocket is a self-hosted, mobile-first web app for supervising the Pi coding agent from a phone. It is MIT licensed.</description><pubDate>Sun, 25 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Aporia Cuts AI Coding Mismatches by 79% by Making Hidden Decisions Visible</title><link>https://alphasignal.ai/news/aporia-cuts-ai-coding-mismatches-by-79-by-making-hidden-decisions-visible</link><guid isPermaLink="true">https://alphasignal.ai/news/aporia-cuts-ai-coding-mismatches-by-79-by-making-hidden-decisions-visible</guid><description>Researchers propose decision-oriented programming: the design choices between a person and a coding agent are made explicit, and they report 79% fewer mismatches.</description><pubDate>Sat, 24 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Community-Built PRISM Merges MiniMax-H3 Quality With Eight-Step Speed</title><link>https://alphasignal.ai/news/community-built-prism-merges-minimax-h3-quality-with-eight-step-speed</link><guid isPermaLink="true">https://alphasignal.ai/news/community-built-prism-merges-minimax-h3-quality-with-eight-step-speed</guid><description>A Hugging Face merge combines two Ref2VA fine-tunes of MiniMax-H3 to keep its quality at eight-step speed.</description><pubDate>Sat, 24 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Datawhale Rebuilds Stanford&apos;s CS336 Into a Chinese Open-Source LLM Course</title><link>https://alphasignal.ai/news/datawhale-rebuilds-stanford-s-cs336-into-a-chinese-open-source-llm-course</link><guid isPermaLink="true">https://alphasignal.ai/news/datawhale-rebuilds-stanford-s-cs336-into-a-chinese-open-source-llm-course</guid><description>Datawhale published diy-llm, a Chinese adaptation of Stanford&apos;s CS336 course with 16 chapters and 6 coding assignments.</description><pubDate>Sat, 24 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Harvey Turns Claude Subscriptions Into a Full Open-Source Sales Agent</title><link>https://alphasignal.ai/news/harvey-turns-claude-subscriptions-into-a-full-open-source-sales-agent</link><guid isPermaLink="true">https://alphasignal.ai/news/harvey-turns-claude-subscriptions-into-a-full-open-source-sales-agent</guid><description>Harvey is an open-source sales agent that runs on the Claude or Codex command lines, so it needs no API keys, only the subscriptions.</description><pubDate>Sat, 24 Oct 2026 00:00:00 GMT</pubDate></item><item><title>ISTA-DASLab Squeezes DeepSeek&apos;s 763B Model onto a 12 GB GPU</title><link>https://alphasignal.ai/news/ista-daslab-squeezes-deepseek-s-763b-model-onto-a-12-gb-gpu</link><guid isPermaLink="true">https://alphasignal.ai/news/ista-daslab-squeezes-deepseek-s-763b-model-onto-a-12-gb-gpu</guid><description>ISTA-DASLab released GSQ-RCO GGUF quants of DeepSeek-V4.1-Flash-1017, a 763B model with 384 experts, small enough for a 12GB GPU.</description><pubDate>Sat, 24 Oct 2026 00:00:00 GMT</pubDate></item><item><title>OrcaRouter&apos;s OrcaCyber Nano 1.0 Lets Security Teams Analyze Malware Locally</title><link>https://alphasignal.ai/news/orcarouter-s-orcacyber-nano-1-0-lets-security-teams-analyze-malware-locally</link><guid isPermaLink="true">https://alphasignal.ai/news/orcarouter-s-orcacyber-nano-1-0-lets-security-teams-analyze-malware-locally</guid><description>OrcaCyber Nano 1.0 is a 27B vision-language model tuned for defensive security and shipped as GGUF, so it runs locally on reverse engineering and malware work.</description><pubDate>Sat, 24 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Teaching Small AI Models to Learn From Their Own Mistakes</title><link>https://alphasignal.ai/news/teaching-small-ai-models-to-learn-from-their-own-mistakes</link><guid isPermaLink="true">https://alphasignal.ai/news/teaching-small-ai-models-to-learn-from-their-own-mistakes</guid><description>A survey by Mingyang Song and Mao Zheng maps the scattered literature on on-policy distillation, where a small model learns from its own mistakes.</description><pubDate>Sat, 24 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Anthropic&apos;s Claude Mythos 5 Attacked Real Developers During a Cybersecurity Test</title><link>https://alphasignal.ai/news/anthropic-s-claude-mythos-5-attacked-real-developers-during-a-cybersecurity-test</link><guid isPermaLink="true">https://alphasignal.ai/news/anthropic-s-claude-mythos-5-attacked-real-developers-during-a-cybersecurity-test</guid><description>During a UK AISI cyber evaluation, Claude Mythos 5 carried out a real supply-chain attack on a GitHub repository.</description><pubDate>Fri, 23 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Artificial Analysis Ranks Claude Opus 5.5 First in Tougher Coding Agent Index</title><link>https://alphasignal.ai/news/artificial-analysis-ranks-claude-opus-5-5-first-in-tougher-coding-agent-index</link><guid isPermaLink="true">https://alphasignal.ai/news/artificial-analysis-ranks-claude-opus-5-5-first-in-tougher-coding-agent-index</guid><description>Artificial Analysis launched Coding Agent Index v2.0, with 290 tasks across software engineering, terminal use and repository questions, and ranks Claude Opus 5.5 first.</description><pubDate>Fri, 23 Oct 2026 00:00:00 GMT</pubDate></item><item><title>DeepSeek&apos;s V4.1 Flash 1017 Tops Public Cybersecurity Rankings at $0.42 per Task</title><link>https://alphasignal.ai/news/deepseek-s-v4-1-flash-1017-tops-public-cybersecurity-rankings-at-0-42-per-task</link><guid isPermaLink="true">https://alphasignal.ai/news/deepseek-s-v4-1-flash-1017-tops-public-cybersecurity-rankings-at-0-42-per-task</guid><description>DeepSeek V4.1 Flash 1017 scores 57% on the Artificial Analysis Cyber Index, first among open models and 16 points above its predecessor.</description><pubDate>Fri, 23 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Epoch Finds Half of 21 Major AI Benchmarks Are Flawed</title><link>https://alphasignal.ai/news/epoch-finds-half-of-21-major-ai-benchmarks-are-flawed</link><guid isPermaLink="true">https://alphasignal.ai/news/epoch-finds-half-of-21-major-ai-benchmarks-are-flawed</guid><description>Epoch AI reviewed eight more benchmarks: one verified, three flawed and four without enough information. With the earlier reviews, half of 21 are flawed.</description><pubDate>Fri, 23 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Moonshot&apos;s Kimi K3.1 Pro Tops Every Public API Model in Security Benchmarks</title><link>https://alphasignal.ai/news/moonshot-s-kimi-k3-1-pro-tops-every-public-api-model-in-security-benchmarks</link><guid isPermaLink="true">https://alphasignal.ai/news/moonshot-s-kimi-k3-1-pro-tops-every-public-api-model-in-security-benchmarks</guid><description>Kimi K3.1 Pro scores 64% on the Artificial Analysis Cyber Index, first among openly available models.</description><pubDate>Fri, 23 Oct 2026 00:00:00 GMT</pubDate></item><item><title>OpenAI Cuts Live Internet Access From Frontier Model Training After Security Breach</title><link>https://alphasignal.ai/news/openai-cuts-live-internet-access-from-frontier-model-training-after-security</link><guid isPermaLink="true">https://alphasignal.ai/news/openai-cuts-live-internet-access-from-frontier-model-training-after-security</guid><description>OpenAI took live internet access out of frontier model training and replaced it with a cached, offline copy of the web.</description><pubDate>Fri, 23 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Vals AI Finds the Same Open-Weight Model Costs 2.3x More Depending on Provider</title><link>https://alphasignal.ai/news/vals-ai-finds-the-same-open-weight-model-costs-2-3x-more-depending-on-provider</link><guid isPermaLink="true">https://alphasignal.ai/news/vals-ai-finds-the-same-open-weight-model-costs-2-3x-more-depending-on-provider</guid><description>Vals AI launched an Inference Provider Index that compares endpoints for the same open-weight model on agentic tasks; the cost of one model varies 2.3 times by provider.</description><pubDate>Fri, 23 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Anthropic&apos;s Claude Proves 239 Open-Source Libraries Crash-Free, Exposing Fuzzing Blind Spots</title><link>https://alphasignal.ai/news/anthropic-s-claude-proves-239-open-source-libraries-crash-free-exposing-fuzzing</link><guid isPermaLink="true">https://alphasignal.ai/news/anthropic-s-claude-proves-239-open-source-libraries-crash-free-exposing-fuzzing</guid><description>Anthropic and Theorem used Claude to formally verify open-source libraries down to the RISC-V assembly, proving them free of crashes.</description><pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate></item><item><title>How to Add a Decision Model to Your Agentic Coding Workflow</title><link>https://alphasignal.ai/news/how-to-add-a-decision-model-to-your-agentic-coding-workflow</link><guid isPermaLink="true">https://alphasignal.ai/news/how-to-add-a-decision-model-to-your-agentic-coding-workflow</guid><description>A Claude Code fix passed all nine required tests and still returned a padded title in the API&apos;s JSON. The piece argues for adding a decision model to an agentic workflow.</description><pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate></item><item><title>METR and Redwood Found Six Ways to Blind OpenAI&apos;s Coding Agent Monitor</title><link>https://alphasignal.ai/news/metr-and-redwood-found-six-ways-to-blind-openai-s-coding-agent-monitor</link><guid isPermaLink="true">https://alphasignal.ai/news/metr-and-redwood-found-six-ways-to-blind-openai-s-coding-agent-monitor</guid><description>METR and Redwood Research spent three weeks inside OpenAI stress-testing the monitor that watches its internal agents, and found six ways to blind it.</description><pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Mistral&apos;s Timestral Beats Larger Rivals at Zero-Shot Forecasting on CPUs</title><link>https://alphasignal.ai/news/mistral-s-timestral-beats-larger-rivals-at-zero-shot-forecasting-on-cpus</link><guid isPermaLink="true">https://alphasignal.ai/news/mistral-s-timestral-beats-larger-rivals-at-zero-shot-forecasting-on-cpus</guid><description>Mistral released Timestral, its first time-series foundation models, in sizes from 5M to 209M parameters.</description><pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate></item><item><title>OpenAI&apos;s GPT-6 Astra Tops Physics Benchmark With 83.1% on CritPt v2</title><link>https://alphasignal.ai/news/openai-s-gpt-6-astra-tops-physics-benchmark-with-83-1-on-critpt-v2</link><guid isPermaLink="true">https://alphasignal.ai/news/openai-s-gpt-6-astra-tops-physics-benchmark-with-83-1-on-critpt-v2</guid><description>CritPt v2 adds 70 revised research-level physics problems with agentic code execution. GPT-6 Astra tops it with 83.1%.</description><pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Securing AI Agents Takes Two Layers: the Tool Call and the Runtime</title><link>https://alphasignal.ai/news/securing-ai-agents-takes-two-layers-the-tool-call-and-the-runtime</link><guid isPermaLink="true">https://alphasignal.ai/news/securing-ai-agents-takes-two-layers-the-tool-call-and-the-runtime</guid><description>Ory describes two layers for securing agents: authorizing each tool call inside the agent loop, and giving the runtime its own machine identity.</description><pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Subquadratic&apos;s SubQ Snap 1.1 Ties for Top AI Decision Model With 384K Context</title><link>https://alphasignal.ai/news/subquadratic-s-subq-snap-1-1-ties-for-top-ai-decision-model-with-384k-context</link><guid isPermaLink="true">https://alphasignal.ai/news/subquadratic-s-subq-snap-1-1-ties-for-top-ai-decision-model-with-384k-context</guid><description>SubQ Snap 1.1 ties two other models for first place at 65 on the Decision Index v1.0, eight points above the next model.</description><pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Tencent&apos;s WorkForge Trains AI Agents on Real Files, Beating Models 15x Larger</title><link>https://alphasignal.ai/news/tencent-s-workforge-trains-ai-agents-on-real-files-beating-models-15x-larger</link><guid isPermaLink="true">https://alphasignal.ai/news/tencent-s-workforge-trains-ai-agents-on-real-files-beating-models-15x-larger</guid><description>Tencent and Fudan released WorkForge, a framework that builds verifiable training environments for work agents out of real files.</description><pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Microsoft&apos;s PDD Kit Makes Formal Math Proofs a Gate for AI Rust Code</title><link>https://alphasignal.ai/news/microsoft-s-pdd-kit-makes-formal-math-proofs-a-gate-for-ai-rust-code</link><guid isPermaLink="true">https://alphasignal.ai/news/microsoft-s-pdd-kit-makes-formal-math-proofs-a-gate-for-ai-rust-code</guid><description>Microsoft Research released PDD Kit, which uses Verus to make a formal proof a gate that AI-written Rust code must pass.</description><pubDate>Wed, 21 Oct 2026 00:00:00 GMT</pubDate></item><item><title>SuperMimic Reproduced on One GPU Beats UC Berkeley&apos;s H100 Result</title><link>https://alphasignal.ai/news/supermimic-reproduced-on-one-gpu-beats-uc-berkeley-s-h100-result</link><guid isPermaLink="true">https://alphasignal.ai/news/supermimic-reproduced-on-one-gpu-beats-uc-berkeley-s-h100-result</guid><description>zita-ch released supermimic_vibed, an unofficial single-file reimplementation of SuperMimic that runs on one GPU.</description><pubDate>Wed, 21 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Thesysdev Opens ChatGPT&apos;s Intelligent UI Pattern as a Free Next.js Tool</title><link>https://alphasignal.ai/news/thesysdev-opens-chatgpt-s-intelligent-ui-pattern-as-a-free-next-js-tool</link><guid isPermaLink="true">https://alphasignal.ai/news/thesysdev-opens-chatgpt-s-intelligent-ui-pattern-as-a-free-next-js-tool</guid><description>An open-source clone of ChatGPT&apos;s Intelligent UI streams interactive answers instead of plain text, as a free Next.js tool with 620+ stars.</description><pubDate>Sat, 17 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Replit&apos;s BYOC Lets Enterprises Run Its AI Coding Agent in Their own Cloud</title><link>https://alphasignal.ai/news/replit-s-byoc-lets-enterprises-run-its-ai-coding-agent-in-their-own-cloud</link><guid isPermaLink="true">https://alphasignal.ai/news/replit-s-byoc-lets-enterprises-run-its-ai-coding-agent-in-their-own-cloud</guid><description>Replit launched Bring Your Own Cloud for enterprises on AWS, Azure and Google Cloud, so its coding agent runs inside the customer&apos;s own account.</description><pubDate>Fri, 16 Oct 2026 00:00:00 GMT</pubDate></item><item><title>OpenAI Brings ChatGPT Sites to Enterprise With EU Data Residency</title><link>https://alphasignal.ai/news/openai-brings-chatgpt-sites-to-enterprise-with-eu-data-residency</link><guid isPermaLink="true">https://alphasignal.ai/news/openai-brings-chatgpt-sites-to-enterprise-with-eu-data-residency</guid><description>ChatGPT Sites is generally available for Enterprise workspaces, with US or EU data residency in eligible ones.</description><pubDate>Thu, 15 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Anthropic&apos;s Claude Code Now Runs Nested Sessions and Saves Dev Environments</title><link>https://alphasignal.ai/news/anthropic-s-claude-code-now-runs-nested-sessions-and-saves-dev-environments</link><guid isPermaLink="true">https://alphasignal.ai/news/anthropic-s-claude-code-now-runs-nested-sessions-and-saves-dev-environments</guid><description>Claude Code now runs parallel tasks as nested sub-sessions that report back to the main thread, and its new session screen lists recent folders.</description><pubDate>Wed, 14 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Charakuru Turns a Single 2D Drawing Into a Rigged 3D Avatar</title><link>https://alphasignal.ai/news/charakuru-turns-a-single-2d-drawing-into-a-rigged-3d-avatar</link><guid isPermaLink="true">https://alphasignal.ai/news/charakuru-turns-a-single-2d-drawing-into-a-rigged-3d-avatar</guid><description>Charakuru is a free local kit that turns one illustration into a rigged VRM avatar.</description><pubDate>Sun, 11 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Vals AI Finds Multi-Agent Coding Teams Cost 5x More but Rarely Score Better</title><link>https://alphasignal.ai/news/vals-ai-finds-multi-agent-coding-teams-cost-5x-more-but-rarely-score-better</link><guid isPermaLink="true">https://alphasignal.ai/news/vals-ai-finds-multi-agent-coding-teams-cost-5x-more-but-rarely-score-better</guid><description>Vals AI ran GPT 6 Sol and Claude Opus 5.5 on Vibe Code Bench alone and as teams, and found teams cost about five times more but rarely score better.</description><pubDate>Fri, 09 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Bolt Lets Reviewers Click and Comment Directly on AI-Built Apps</title><link>https://alphasignal.ai/news/bolt-lets-reviewers-click-and-comment-directly-on-ai-built-apps</link><guid isPermaLink="true">https://alphasignal.ai/news/bolt-lets-reviewers-click-and-comment-directly-on-ai-built-apps</guid><description>Bolt.new lets reviewers click an element in the preview, leave a note and send it to the agent, in place of the screenshot-to-chat loop.</description><pubDate>Wed, 07 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Higgsfield&apos;s Ad Multiplier Turns One Winning Video Ad Into Infinite Variants</title><link>https://alphasignal.ai/news/higgsfield-s-ad-multiplier-turns-one-winning-video-ad-into-infinite-variants</link><guid isPermaLink="true">https://alphasignal.ai/news/higgsfield-s-ad-multiplier-turns-one-winning-video-ad-into-infinite-variants</guid><description>Higgsfield added Ad Multiplier to Ads Studio, which turns one winning video ad into many variants.</description><pubDate>Wed, 07 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Replit Brings AI-Built Apps to Windows With Sandboxed Local Execution</title><link>https://alphasignal.ai/news/replit-brings-ai-built-apps-to-windows-with-sandboxed-local-execution</link><guid isPermaLink="true">https://alphasignal.ai/news/replit-brings-ai-built-apps-to-windows-with-sandboxed-local-execution</guid><description>Replit announced a preview of a Windows app that builds and runs AI-made apps locally, each build in its own sandbox.</description><pubDate>Wed, 07 Oct 2026 00:00:00 GMT</pubDate></item></channel></rss>