14 AI stories explainx.ai reported on October 8, 2026 so far, ranked by reader interest and grouped by topic. Each links to the full write-up with sources.
Microsoft announced a 3-bit local build of its 137B MAI-Code-1.1 Flash coding model for high-memory Windows PCs, with GitHub Copilot handing work to it; it needs roughly 75GB of memory at full context.
OpenAI confirmed it fired three safety researchers for mishandling sensitive information reportedly shared with an outside AI-safety group; the recipient is unnamed and key details are unpublished.
Microsoft says Windows 11 will route AI work between local hardware and the cloud, with Execution Containers for agents now generally available and local Copilot features arriving over the coming months.
The Association for Human Mathematics, in a statement reposted by Terence Tao on October 7, 2026, urged mathematicians to stop working with OpenAI after its October 6 release of 700-plus math manuscripts.
SemiAnalysis counts over 24 GW operating in China plus about 50 GW in the pipeline versus about 56 GW in the US; the buildout is gated by chips rather than power, so GW overstates effective compute.
Scott Aaronson cites anonymous sources saying AI labs are testing whether models can break cryptography; no break has been reported, and Ethereum researchers are urging precaution, not panic.
Guardian Australia reportedly found OpenAI used AI to help draft its September 10 breach notice, though Jason Kwon told an inquiry he did not believe so; a source says humans reviewed it before sending.
Netflix's October 12 documentary says OpenAI agents hit Hugging Face; the published record confirms the core facts but differs on duration, who built the agents, and how many were involved.
The 81% figure measures task exposure to AI plus robots, not jobs replaced; Anthropic's robot study finds robots cost-competitive for only about 0.3% of tasks, so hiring is not a contradiction.
Microsoft announced on October 7, 2026 that Meta Muse is coming to Windows as a native app with MXC sandbox integration, but no release date, requirements or Windows-specific pricing have been given.
Tab is a text-message personal assistant with its own phone number, computer and wallet that emerged from stealth at a $300M valuation; its privacy promises are claims, and the amount raised and pricing were not disclosed.
Sonnet 5.5 cache reads dropped from $0.20 to $0.10 per million tokens; the roughly 20% saving holds only when cache reads dominate tokens and output is tiny, while chat-style mixes save about 5%.
Perplexity's pplx-embed-v2-late keeps token-level vectors for text and images so parts of a query match parts of a document, in 0.6B and 9B sizes that share an embedding space; benchmark scores were not independently confirmed.
OpenAI says teens average under 15 minutes a day on ChatGPT for Teens and announced College Planner, flashcards and a teen council; the figures are self-reported and landed the same day as a critical outside test.
Max 5x gets $100, Max 20x gets $200 and Team gets $20 per Standard seat or $100 per Premium seat (pooled, capped at $500) in monthly Claude API credits, but they cannot be spent on interactive Claude Code.
AI readiness means ten skills: fluency, context engineering, agent skills and MCP, building with coding agents, evaluation, workflow design, data basics, safety, tool selection and human judgment, practiced on real work.
Paul Graham argues that any business banning AI agents invites a competitor that allows them; the claim is plausible but rests on Amazon having banned agents broadly, when verified actions so far target specific agents.
RunningTab (arXiv 2610.10444) adds an environment-side record of task requirements and read files, so agents working over file workspaces stop dropping content they already saw.
Haiku 5.5, GPT-6 Luna and GLM-5.3-Flash cluster near $0.10-$0.15 input, DeepSeek V4.1 Flash wins on cache price and open weights, and Gemini 3.7 Flash costs about 7x more; cost per task decides, not sticker.
The $2,599.99 Surface Laptop Ultra has 24GB of memory and cannot run the models Microsoft is promoting; the 128GB tier, priced up to about $5,899.99, is what a 137B local coding model needs.
Ten decision models compared: hosted Jev, Perplexity, OpenAI, Cloudflare and Liquid APIs plus open-weight Strands, Security-One, GLiNER2.5-Decide, Julia 1, Kev and Jeff, with vendor numbers flagged as unverified.
+36 more updated posts
AI platform for deep video understanding.
Open-source collaborative agentic workspace for data teams.
Build, personalize, and optimize landing pages with AI.
The Multiplayer AI Office where your team and agents work together.
IrisGo is an AI operating system for workflow automation on Windows and Mac.
Get each day's AI news in your feed reader: daily RSS · every post