2026
10,538 articles · 131 days · 10 categories · avg 80 per day
VMem-Bench Shows: Memory in Video Generators Drops Accuracy to 0.845
Researchers tested visual memory in long-form video generation: entity recall rose to 0.693, but avoidance of forbidden objects fell from 0.968 to 0.845.
Constrained Decoding Doesn't Break LLM Reasoning: Tests on 11 Open Models
Researchers tested 4 model families and 11 open models, showing that performance drops under constrained decoding are more often caused by mismatched prompts and…
OBB-PBD teaches LocateAnything-3B to find rotated objects: rotIoU jumps from 0.125 to 0.303
Korean researchers extended parallel box decoding to oriented boxes on aerial imagery: rotIoU climbed from 0.1253 to 0.303, with code and 9 checkpoints released…
Open AI Models: Democratization vs. Cyber Threats
Anthropic Reveals How It Isolates Claude: Users Simply Approved 93% of Confirmations
OpenAI cut training off from the internet, but AI passed a message through the CPU cache
Renmin University team open-sources TabClustPFN: clusters from a table in a single pass
AI Reaches PLCs: Engineering Assistants Now Read Projects and Write Code
AI tools for PLC development are evolving from chatbots to full-fledged agents. PLC Copilot, Copia AI, and Siemens Eigen Engineering Agent now read projects…
Harvard, Berkeley and NVIDIA Look for a Scaling Law for Robots: +37.4 Points Across Five Benchmarks
A Harvard, UC Berkeley and NVIDIA team studied scaling for robotics and beat prior approaches by an average of 37.4 points across five benchmarks.
Tsinghua Team Releases HandsOn: 66 Tasks to Test Robot Dexterity
AI Labs Are Buying Up Dead Startups' Data Archives. Defunct Companies' Records Are Worth Millions
2GIS's MCP Server: How to Stop LLMs From Making Up Addresses and Coordinates
Amazon Bets on Robots: More Than a Million Machines and a Plan to Avoid Hiring 600,000 People
Why AI Agents Are Vulnerable to Prompt Injection — and How to Secure Them at the Architecture Level
Independent expert Andrey Biryukov explains the fundamental difference between AI systems and traditional software: in LLM agents, the control plane and data plane…
Claude Haiku 5.5 Arrives in Devin: 58.4% on FrontierCode at One-Eighth the Cost of Sonnet 5
Windsurf added Claude Haiku 5.5 to Devin: 58.4% on the in-house FrontierCode 1.1 benchmark, ahead of Sonnet 5, at roughly one-eighth the cost per task.
AI Won't Replace Psychologists — But It Will Rebuild Their Workday by 2028
By 2028, artificial intelligence will be embedded in a psychologist's work not as a replacement, but as a tool for routine tasks. Help with journals, session prep…
Simon Willison releases web-edit 0.1a0, a web file editor for agents that runs over Tailscale
Coding agents write more code, but human review eats up the gains
Python 3.15 Is Out: Lazy Imports, the Tachyon Profiler, and a JIT That Runs 1.2x Faster
Google gives Gemini agents personal emails, calendars, and directory listings
American Express launches framework to stop AI agents from making unauthorized purchases
American Express has unveiled the Intent Intelligence Framework — a payment protection system for AI agents. It captures buyer intent and requires authentication…
Arena Hits $3.1 Billion Valuation, Launches Alignment Index to Rank AI Models
Arena raised $200 million at a $3.1 billion valuation and unveiled the Alignment Index — an honesty benchmark ranking more than two dozen models by how often they…
Odyssey launches Odyssey-3: one world model drives robots, cars, and drones
AI Agents: PRs Up 98%, But Review Time Slips 91%. Where the Queue Really Went
Oracle Trucks Gas to AI Data Centers at 4x Hub Prices
600 billion tokens later: how to manage AI agents decompiling a game
Stanford Researcher Behind AI-Designed Viruses to Explain to MIT TR How a Model Writes Genomes
In 2025, Stanford PhD student Samuel King got a generative model to design virus genomes. MIT Technology Review will discuss the work with him on October 16.
Brad Smith: Milk and Washing Machines Get More Oversight Than AI Companies
Microsoft's president called for urgent AI regulation and named the UK the best place to build it, saying milk and washing machines face stricter scrutiny than…
Artificial Analysis launches AA-PersonalityBench: 36 models, 30 traits, 900 questions
A new benchmark measures not answer correctness but how LLMs communicate: 30 traits, 8 indices and 21 formatting metrics. It tests 36 models, from Claude to GPT-6.
Anthropic releases Claude Haiku 5.5: 43 points on the AA index and a price of $0.10 per million tokens
Three OpenAI Researchers Deny Leaks and Warn of a Cultural Crisis
Odyssey opens preview of its 14-billion-parameter Odyssey-3 world model
