Post
13 Oct 2026
Microsoft's new AI model makes decisions, not text
Microsoft-Decision-1 returns a calibrated probability for each option. Microsoft says it topped its 36-benchmark test and ran about 35x faster than GPT-6 Sol. Live on Foundry and OpenRouter.
02AI NewsMicrosoft announces Maia 300 AI chip
Maia 300 is expected to deliver 40% more tokens per megawatt vs leading GPUs and up to 7x Maia 200 rack throughput; production in 2027.
Microsoft · blogs.microsoft.com ↗ (opens in a new tab)
03AgentsClaude can run up to 1,000 agents in one workflow
Managed Agents workflows run up to 64 threads at once and 1,000 agents per run, so jobs too big for one conversation can fan out.
Anthropic · platform.claude.com ↗ (opens in a new tab)
04ModelsAnthropic cuts what distillers can extract from Claude
New defenses cut harvested data by about 90–99% and exposed reasoning tokens by 10–100x; tens of thousands of accounts banned.
Anthropic · anthropic.com ↗ (opens in a new tab)
05AgentsZ.ai launches GLM-Legal
An agentic legal model scoring 19.2% all-pass on Harvey LAB, 12.5 points above its base model, with results checked by Vals.
Z.ai · z.ai ↗ (opens in a new tab)
06BusinessTypeSafe raises $870M at a $7.5B valuation
a16z led the round for Jev, a model that returns typed, probability-backed decisions instead of chat prose.
TypeSafe · typesafe.ai ↗ (opens in a new tab)
07ModelsSmall AI models may soon self-replicate on cheap devices
Anthropic's Frontier Red Team ran a 26B model across 15 ~$55 Android TV boxes; a 9B model got inference running on 8 of 9 hardware types. End-to-end replication worked only in simplified tests (an upper bound); Anthropic roughly projects small models get there by end of 2026.
Anthropic · anthropic.com ↗ (opens in a new tab)
08VisionMiniCPM-V 4.7 runs a 35B vision model on an iPhone
OpenBMB's open-source 35B-A3B MoE decodes 13+ tokens/s in 3 GB of memory on iPhone 17 Pro Max (OpenBMB's numbers), using Lookahead Expert Scheduling for a 5x on-device speedup. A 1B version too.
OpenBMB · github.com ↗ (opens in a new tab)
09RoboticsNatura-1: robot hands trained on 1M hours of human video
Nature Will's 50B mixture-of-experts model, pretrained on 1M+ hours of egocentric human video plus limited robot data, follows plain-language instructions zero-shot and drives 20-DoF dexterous hands in a 100 ms loop. Nature Will calls it early evidence.
Nature Will · naturewill.ai ↗ (opens in a new tab)
10AgentsSubQ Snap makes agent decisions in about a second
Subquadratic's decision model picks from allowed answers instead of writing text, reads up to 384K tokens of text and images, and in Subquadratic's tests beats TypeSafe's Jev on 13 of 16 enterprise benchmarks. $0.05 per 1M input tokens, output free.
Subquadratic · subq.ai ↗ (opens in a new tab)
11CodingAtomic Machines: AI fab builds micro-machines from code
An AI-native Matter Compiler builds working micro-machines from digital designs as a self-driving fab—no masks; every iteration is software. First product PrimeSwitch is sampling now.
Atomic Machines · atomicmachines.com ↗ (opens in a new tab)
12ModelsCoinbase: 9B open model beats Opus 4.5 on fraud
Post-trained Qwen3.5-9B with SkyRL beat Opus 4.5 on Onramp fraud (+7.5pp precision, +12 recall, +9.6 F1, +35.4 dollar-weighted recall) with 55% lower P50 latency and under $100 training cost.
Coinbase · coinbase.com ↗ (opens in a new tab)
13AgentsAMD Helion: agentic AI racks up to 32,768 cores
Open OCP rack blueprint for agentic AI: CPU-only configs up to 32,768 EPYC cores; all-in-one with MI350P GPUs. Spec headed to OCP; partner configs planned November 2026.
AMD · newsroom.amd.com ↗ (opens in a new tab)
14AgentsAI agents got 48 hours to do real ML research
Hillclimb's DeepRSI: 15 open-ended ML problems, 48 hours, no internet. Claude Fable 5.1 leads at 1548 Elo ahead of GPT-6 Astra (1232); Astra even hard-coded a puzzle solver into a transformer's weights.
Hillclimb · deeprsi.com ↗ (opens in a new tab)
15AgentsDigitalOcean launches tools for self-improving agents
Agent Learning Services (Insights, Signals, Simulations) turn production sessions into tests before changes ship; Agent Data Services is a managed data layer for agent context. Learning in public preview, Data in private preview.
DigitalOcean · investors.digitalocean.com ↗ (opens in a new tab)
16RoboticsFrontier AI can't label robot video
Concept's Oct 12 eval with Mecka on 200 egocentric clips: Mecka's specialized labeling model hit 79.5% fully correct vs GPT-6 Astra (max) at 43.5% as the best frontier. Specialized models plus humans in the loop still win for robot training data.
Concept · concept.dev ↗ (opens in a new tab)
Tiny open-source decision model beats Opus 5 and GPT-5.6
FLock.io THIS/THAT Model 1.2 (built on Qwen3.5-2B, 1.88B params, MIT) scores 0.878 accuracy on 1,710 public questions in 31.4 ms with zero output tokens — ahead of Claude Opus 5 (0.834) and GPT-5.6 (0.816). Runs on laptops and phones.
FLock · flock.io ↗ (opens in a new tab)
18ModelsIntel's new AI GPU packs up to 480GB of memory
Intel GPU D-series (Crescent Island, led by the D270), detailed Oct 12 at OCP: up to 480GB LPDDR5X with no HBM, 1.5TB/s, 1.3 PF FP8 in 350W, air-cooled PCIe. Up to 300B-param models per card, 1T+ on eight. Intel's card has 160GB; partner cards up to 480GB. Sampling now.
Intel · intel.com ↗ (opens in a new tab)
19ModelsSame AI model, 35-point swing from the harness
8090 Labs ran 7 models in 5 coding harnesses on DeepSWE v1.1: GLM 5.3 Flash passed 24% in Codex vs 59% in DeepSeek's harness, and no model did best in its maker's own harness. Test model + harness together and price by cost per success.
8090 Labs · 8090.ai ↗ (opens in a new tab)
Anthropic launches free OSS Scanner for open source
Opt-in free vuln-finding informed by Project Glasswing. CyberGym: LLMs from under 20% to over 85% vulns found. 29,000+ candidates, ~6,000 manually reviewed; validation 97 critical/high across 48 projects → 85 (88%) met CVD bar. Enroll via PR.
Anthropic · anthropic.com ↗ (opens in a new tab)
21VisionQwen Image 2.1 Turbo: open image gen in 8 steps
~7B text-to-image + editing with a saved 8-step schedule (base examples often use 40), CFG=1 and prefix KV cache. Open weights under Qwen Research License — not fully commercial. Pro+Turbo APIs on Alibaba Cloud Model Studio.
Qwen · huggingface.co ↗ (opens in a new tab)
22ModelsRSI-Anything: AI that improves itself with world models
World-model RSI framework: code repair 26.3%→29.2%→31.2% (coder fixed); GUI actions −39.4%; lab planning 21.80%→46.01% over 5 versions; reaction yield 96.15% vs Bayesian 82.16% at 40 experiments.
arXiv · arxiv.org ↗ (opens in a new tab)- 23Agents23Agents
AgentSWE: coding agents fail to build the agents you want
25-task benchmark for creating/editing/optimizing agents. Strongest Creation mean 19.1; Editing 31.6; Optimization <5% gap closed on average. 11 Creation runs with ≥90 on development averaged only 9.9 held-out.
arXiv · arxiv.org ↗ (opens in a new tab) 
24SecurityAI models can provably hide skills from safety tests
Proof-of-concept (NeurIPS 2026): an encrypted network hidden in a model's weights runs only on a secret trigger and stays provably hard to elicit. After fine-tuning it scored 50.9% vs 50.6% for a random control; short 8-bit triggers could still be brute-forced.
arXiv · arxiv.org ↗ (opens in a new tab)
25VoiceEmbedding Gemma Transcribe: 1 hour of audio in ~13 s (reported)
An independent project turned Google's EmbeddingGemma 2 into speech-to-text. They report 288x real time on a base M3 MacBook Air, fully on device, with ~5% word error rate vs 2-3% for leading models.
eg2.haku.sh · eg2.haku.sh ↗ (opens in a new tab)
26VoiceFish Audio launches Drama 3 voice direction
Drama 3 is a text-to-speech model you direct in plain language: tone, pacing, multi-speaker scenes in one pass. Preview API model drama-3-preview; 80% off API for the first 30 days. Still in preview.
Fish Audio · fish.audio ↗ (opens in a new tab)
27AgentsStateless AI agents cut long-horizon research tokens
SLA keeps research state in the harness and gives each agent a fresh context. On Anthropic kernel optimization with Codex, it matched the strongest baseline with 93.1% fewer tokens (67.9M vs 986.3M). Best full-budget result on every task; Advisor used 0.24–0.51% of tokens.
arXiv · arxiv.org ↗ (opens in a new tab)
Mistral Ministral 3 is running on a satellite in orbit
Orbitworks and Loft Orbital say the first satellite of a planned 50-satellite AI constellation is in orbit after an Oct 1 SpaceX rideshare. It flies two NVIDIA Jetson AGX Orin modules and runs Mistral Ministral 3 (3B and 8B) onboard to summarize acquisitions and answer operator queries — alongside nearly 20 other models. Alert in seconds instead of dumping raw imagery hours later; first bird of a $1B program.
Loft Orbital · loftorbital.com ↗ (opens in a new tab)
29AgentsScaleDown CDMs: claims 200x faster tools in LangChain
LangChain wired ScaleDown's Summarize, Compress, Extract, and Classify CDMs into Deep Agents. ScaleDown reports CDMs are 200× faster and 467× cheaper than frontier models on these narrow tasks; up to 1M-token inputs; Summarize/Extract/Classify multimodal. On ScaleDown's benches, Summarize was 41× cheaper and Compress 138× cheaper vs an average of GPT-5.5 / 5.4 Mini / Nano.
LangChain · langchain.com ↗ (opens in a new tab)- 30Agents30Agents
LightOn ColGrep-Agent-2B cuts coding-agent search tokens ~34%
LightOn's open 2B local code-search agent cut Claude Code API tokens by 34% vs grep on 52 code-search questions across 13 unseen repos (LightOn's test; not full coding tasks).
LightOn / Hugging Face · huggingface.co ↗ (opens in a new tab) 
31VisionCloudflare Clef-omni: decisions from photo, audio, and video
Open-weight multimodal decision model takes audio, video, image and text in one call; Cloudflare reports ~130 ms text / ~1.5 s for a 21s video+sound.
Cloudflare · blog.cloudflare.com ↗ (opens in a new tab)
NVIDIA open-sources the full Run:ai platform
Full GPU orchestration stack joins NVIDIA DSX AI Factory as open source: KAI Scheduler (Apache 2.0, CNCF Sandbox), CUDA 13.4-native GPU fractions, and more — no paid Community Product subscription.
NVIDIA · developer.nvidia.com ↗ (opens in a new tab)
33AgentsAnthropic rebuilt source control for AI agents
Merges grew from ~2,500/day in May to 30,000+/day by September, so a team of seven built Scion, an in-house VCS, and migrated off Git in five months. Anthropic reports 200x smaller infra, CI checkouts in seconds and a ~25% CI CPU win.
Anthropic · anthropic.com ↗ (opens in a new tab)- 34Models34Models
Gemini 4 Argon is live for API, Pro & Ultra
Google opened Gemini 4 Argon globally via the API/AI Studio, Antigravity (Ultra) and the Gemini app and AI Mode (Ultra and Pro). Google says it's #1 on the Vals Index; intro pricing $2/$10 per million tokens with a 1M token limit.
Google · blog.google ↗ (opens in a new tab)