AI Daily · Oct 18, 2026: OpenCode breach from distillation attack hits 1.4% of users
OpenCode's Console was breached in a large-scale distillation attack affecting 1.4% of users with linked external providers; meanwhile DeepSeek V4.1 Flash 1017 landed on Ollama Cloud and OpenRouter, and Requesty listed DeepSeek V4.1 at 51% off with cached-token pricing.
Pick a date
-
OpenCode reports breach from distillation attack hitting 1.4% of users
OpenCode said a large-scale distillation attack breached its Console, affecting 1.4% of users who had external providers linked; those users were emailed and the team is working with Anthropic, OpenAI and Google.
-
DeepSeek V4.1 Flash 1017 launches on Ollama Cloud and OpenRouter
DeepSeek-V4.1-Flash-1017 is available on Ollama Cloud (US/EU hosting, zero data retention) and OpenRouter, with stronger benchmarks, shorter outputs, and lower cost per task than the prior version; existing deepseek-v4.1-flash users are being migrated.
-
Requesty offers DeepSeek V4.1 at 51% discount with cached token pricing
Requesty lists DeepSeek V4.1 at a 51% discount: $0.098 per 1M input tokens, $0.294 per 1M output tokens, and $0.003 per 1M cached tokens.
-
OpenAI Codex rate limit reset scheduled for day 13
An OpenAI Codex staff member said the day-13 rate limit reset is landing by end of day PST.
-
Command Code offers free models including Cat Jam, Beam 1.0, Glyph Cluster, Laguna S 2.1
Command Code is offering free access to four models — Cat Jam, Beam 1.0, Glyph Cluster and Laguna S 2.1 — plus free tokens for users to build and ship.
-
OpenRouter publishes cookbooks for ScaleDown task models at 20% off
OpenRouter released two runnable cookbooks for ScaleDown's task-specific models, currently 20% off, covering agent-session summarization and a Decisions API that detects stuck agent loops.