#LLM
Curated coverage on «LLM»: releases, research, benchmarks, open-source.
14 AI models spot OLX scammers, but 37 of 42 quietly help with prepayment
A Kazakhstan developer tested 14 LLMs on 72 scam scenarios. Classification scores ranged from 0.93 to 1.00, but only 5 of 42 OLX prepayment replies included a…
Cloudflare Launches Auto Router: It Picks the Model So Employees Don't Have To, Saving Up to 30%
Cloudflare put Auto Router for AI Gateway into beta. In an internal test it cost $2.10 vs. $5.91 for Claude Opus 5.5, with an 86.6% task success rate vs. 96.6%.
T-Bank replaces a swarm of LLMs with a single 32B model and cuts query costs by 2.8–3.9x
At EMNLP 2026, T-Bank researchers showed how one Qwen3-32B-based model covers 200+ services and beats Qwen3-235B on corporate benchmarks: 69.6 vs 65.8.
Anthropic files S-1: half-year revenue of $16.3 billion, with a valuation of up to $2 trillion expected
Anthropic Urges Teams to Fold Subagents Back In: A Single Loop With Skills Beats Orchestrators
A ~/.claude Architecture With 200+ Skills: Keeping the System Prompt Under 50K Tokens and Paying $0.15 Per Session
Cloudflare rebuilds Containers for agents: startup is 6x faster
ElevenLabs Hands Its Support Desk to an AI Agent: 22,000 Tickets a Month, 65% Without Humans
ElevenLabs runs its own support on ElevenAgents: the agent fully resolves 65% of requests, replies in 60 seconds, and 70% of its conversations get positive ratings.
Anthropic Advises Cutting Half Your Subagents: One Loop Ran Twice as Fast
Anthropic's Applied AI team reviewed enterprise deployments: a single agent with skills often beats a multi-agent harness on quality, cost and speed. One client saw…
Cloudflare cuts the cost of AI code review by 68%, from $2.43 to $0.77
Fudan and Huawei Xiaoyi Teach an AI Agent to Write Operators for Ascend NPUs: 95.5% Success Rate
Voice Input in China Is Learning to Write for You: From Dictation to Finished Text
EmbodiedLab: robots tested on experiments, and the best one solves less than half
OpenAI Says Its AI Disproved a 30-Year-Old Theoretical Computer Science Conjecture
On October 26, OpenAI announced the disproof of a conjecture that had been considered true for nearly 30 years. Its original authors are now organizing colleagues…
A 12-Author Review Charts How to Trust AI That Improves Itself
Researchers broke down recursive self-improvement in AI: four trust themes, three lenses of verification, and S1-S4 stages where quality gains can easily turn out…