The A.I. Daily · Ten minutes a day to keep up with AIResearchers exploited ChatGPT's HTML Preview and sharing features to build a malicious shared conversation; a victim only had to open the link for attacker-controlled JavaScript to run in their signed-in session, calling authenticated APIs, reading private chats and exfiltrating data. OpenAI fixed the issue on October 26 and rated it P1.
Hacker NewsDetails →
DeepSeek releases enhanced coding agent to rival AnthropicDeepSeek has released a prototype of a more autonomous coding agent, aiming to challenge Anthropic's popular Claude Code. The move shows DeepSeek extending from model development into agent products.
Discussing common patterns in ML techniques across modalitiesA Reddit user initiated a discussion on recurring ML techniques across text, image, and audio modalities (e.g., tokenization, masking, diffusion), asking whether taxonomies exist that group techniques by underlying pattern rather than modality.
AI smart glasses can answer 80% of SAT questions, study findsGerman researchers simulated distorted images from smart glasses to test AI models on SAT questions. Gemini 1.5 Pro-002 scored 82.78% accuracy, GPT-4o 75.69%, raising fairness concerns for exams.
Five Qwen3.8-27B models beat Fable 5 on LiveCodeBench HardUsing the GVS5H method, five Qwen3.8-27B models collaborated via a Memdoor multi-agent workflow, achieving a mean score of 9.8/10 over five passes on the first ten LiveCodeBench Hard problems, surpassing Fable 5's standalone 8.0.
OpenAI may merge ChatGPT and Codex usage limitsA Hacker News discussion indicates OpenAI is merging usage limits for ChatGPT and Codex. If implemented, how users access coding models under one subscription would change.
EU, Canada forge strategic alliance on minerals, AIAt the Montreal summit, the EU and Canada announced deeper cooperation on critical minerals, AI and digital trade, calling it the bloc's closest partnership with a non-European state. European supercomputing infrastructure will open to Canadian researchers, and the European Investment Bank can finance Canadian critical mineral projects.
Less than three years ago, GPT-4 was the best model in the world, as it had been for a year, and we were all speculating whether there was some unique secret method that OpenAI had or whether GPT-4 was just as good as LLMs would ever get. (The answer to both…
The Pro series of GPT models continues to be really impressive for one-shot problems that involve complex calculations, references, or sophisticated ideas. It isn't particularly fast (or particularly pleasant) but GPT-6 Pro gives that "ask one question of the…
👀🗣️👂 [Quoting @DotyBilly]: $meta muse voice is soooo good. It quietly rolled out on 10/24 apparently. Like I think in the future everyone will be talking to a voice agent while performing work tasks.
join me all the way on the light side guys low effort + critique loop > superultramaxhighwhatever (for the throughput-constrained; see swyxio/skills/babysit-runs) [Quoting @theo]: Julius convinced me to move from extra high to high on Opus 5.5 a week or tw…
Dopamine does not spike highest for guaranteed rewards. It spikes highest at reward prediction error. If you want to have fun in life, you should stop avoiding risk and uncertainty.
throw a brick in SF and you’ll hit: 50 inference providers 80 RL environment startups 4090 dudes brokering GPUs
Boot Sector LLMA tiny char-level neural network in a 510-byte MBR that endlessly generates Linux boot logs.
SKIVoice-first coding harness to run multiple coding agents side by side on your Mac.
GitHub Copilot FusionGitHub AI coding agent that routes tasks to suitable models and combines them
BaseTab AITurns the browser new tab into an AI-powered customizable workspace dashboard
BackenlyAn autonomous, self-healing backend platform built for AI coding agents.
Coding WrappedTurns local Claude Code and Codex history into a pixel-art Wrapped report
telemetry.devLLM observability platform that traces each AI agent run, cost and errors.
FlashAttention SQL DemoAutomatically deduce a FlashAttention kernel from ~25 lines of SQL
Gives every coding agent persistent, AI-compressed context across sessions
Design vocabulary for coding agents to catch AI slop before shipping
Open-source Grok Bot alternative with a VM that bots can use
Local-first open-source AI team app where a commander coordinates 16 agents
A headless browser built from scratch in Zig for AI agents and automation
A collection of agent skills for real engineers from one .agents directory
Turns Claude Code into a game studio with 49 AI agents and 72 skills
An agentic skills framework and software development methodology that works
Open multi-agent interactive classroom offering immersive learning in one click
OpenAI cut its funding target from $30B to $20B and its valuation from $1.4T a month ago…
