Author
OpenCode
By and about OpenCode
OpenCode Console breach: exposed provider keys, user losses, and refund updates
OpenCode says a breach affected 1.4% of Console users. Follow-ups clarify the scope, Anthropic key revocations, continued attack attempts, and OpenAI refunds.
blogAI Socratic October 2026 — The Observability Effect
GPT-6 and Claude 5.5, Gemini 4 Argon, Mistral and Beam previews, agent protocols, new tools, AI-generated mathematics, and virtual biology: September 22–October 8, 2026.
newsMulti-harness RL improves a small model across four agent interfaces
Hugging Face and Liquid AI train a 2.6B model across four agent harnesses, raising held-out task success from 42.2% to 54.2% while reducing tool calls.
newsThe week's top AI papers say the harness, not the model, is the variable
DAIR.AI's ten-paper roundup shows scaffolding, not the model, drives the gains: Prime Intellect's Prime Agent lifts ARC-AGI-3 Best@1 from 30% to 95.5%, while a compaction bug quietly erases 90% of safety rules after five rounds.
newsZ.ai's GLM-5.3-Flash: two cheap attentions, MIT weights, and a claim it all ran on Chinese chips
Z.ai revealed OpenRouter's anonymous ox-alpha as GLM-5.3-Flash, a 320B-A18B MIT-licensed MoE that pairs sparse and linear attention to cut KV cache 4.44x, scores 57 on Artificial Analysis's index, and it says was served entirely on Chinese chips.
newsZ.ai says it served GLM-5.3-Flash, OpenRouter's anonymous ox-alpha, entirely on Chinese chips
Z.ai revealed that ox-alpha, the anonymous model that led OpenRouter for six days with 23.2T tokens, is GLM-5.3-Flash, and says it was served entirely on 100,000+ domestic Chinese chips at $0.15/$0.50 per million tokens.
newsNVIDIA finds skill doc-scans predict nothing about what a skill does at runtime
NVIDIA researchers' ACES framework ran 947 paired agent trials across 58 production skills and found that both structural and LLM-judge doc-scan scores correlate with real runtime skill lift at essentially zero (Spearman ρ = -0.018 and -0.027).
newsAgents that post-train AI revise their strategy 2% of the time
A Tsinghua, Renmin and UESTC study of 1,338 AI agent transcripts post-training language models found agents changed strategy in just 2.1% of 3,557 consecutive run pairs, even as execution gains lifted benchmark scores from 10.41% to 23.0%.
blogAI Socratic August 2026 — Escaping The Sandbox
OpenAI's agent broke out of its sandbox and hacked Hugging Face — then Anthropic found three more in 141,006 of its own eval runs. Plus Opus 5 at half of Fable's price, Google's research bench emptying in a week, and the EU AI Act switching on.
newsCoding agents are learning to talk to each other
Anthropic's cross-session messaging for Claude Code, Agent-Manager's tmux orchestration, and Matt Pocock's agent-spawning patterns show developers building multi-agent coordination directly into coding tools rather than bolting it on from outside.
blogAI Socratic July 2025 — The CLI War
The most important AI news and updates from June 15 to July 15.
newsLet's enter the AI Code CLI war
A comparison of six AI code CLI tools—claude-code, anon-kode, codex, opencode, ampcode, and gemini—as Google and AWS enter the command-line coding market.