llm-anthropic 0.31 adds Anthropic Fable 5.2 and Mythos 5.2
Willison says max reasoning on Fable 5.2 spent 128,000 tokens thinking and returned no image, a bug he also saw on Opus 5.5.
34 stories, newest first · one card per story
Willison says max reasoning on Fable 5.2 spent 128,000 tokens thinking and returned no image, a bug he also saw on Opus 5.5.
CLI argument masking by position leaves secrets like --conn-extra and Variable descriptions in cleartext audit logs.
DART improved scores by up to 9.0 points and cut MCP cost by up to 27.1k tokens per task.
The feed only records 47 points and 27 comments, with no release notes or feature list.

Poisoned ovhcloud-cli releases were online for 1 hour 33 minutes and downloaded about 250 times.

Tampered builds v0.11.0 through v0.15.0 were live for under two hours on October 19; v0.16.0 is clean.

Releases v0.11.0 through v0.15.0 were replaced for about 94 minutes on October 19; v0.16.0 is clean.

A hijacked GitHub account pushed trojanized binaries for about 90 minutes on October 19; 0.10.0 and 0.16.0 are clean.

Compromised builds v0.11.0 through v0.15.0 were online for 94 minutes on October 19 before OVHcloud blocked them.

Malicious builds of v0.11.0 through v0.15.0 were blocked within 94 minutes; v0.16.0 is the clean replacement.

The swapped archives were live for up to 94 minutes on October 19 and only reached unhashed installs on selected platforms.
sslx is pitched as a faster, modern command-line alternative to OpenSSL.
A GitHub issue reports OpenAI Codex CLI installs a persistent, self-updating background daemon by default.
Simon Willison released ttok 0.4, a CLI token counter built on OpenAI's tiktoken library.

Adversa AI shows encrypted prompts can make GitHub Copilot CLI read local secrets and exfiltrate them.
New Relic released Ground Truth CLI so developers and AI agents can investigate incidents from the terminal.
Together AI launched Together Link, a free CLI that routes coding agents to open models.
YARA-X 1.21.0 adds five improvements and four bugfixes, including stdin support for --scan-list.
KaliBench shows no open-weight model exceeds 42% exact-command accuracy generating Kali Linux CLI commands without tool hints.
KaliBench measures whether LLMs can turn analyst requests into exact Kali Linux commands across 1,642 tools.
HybridCUA-9B combines GUI and CLI actions, reaching 53.6% on OSWorld, up 14.8 points.

Cloudflare launched cf, an agent-oriented CLI that exposes more than 3,000 API operations with JSON by default.

Cloudflare open-sources Forge, a pipeline that generates SDKs, CLIs, docs, and agent-facing API surfaces.
commit-rewriter 0.2 adds support for rewriting commits on non-default git branches.
Today 1.5K readers · 2.4K views · 754 reads · all time 189K views