For people who build with AI: coding agents and dev tools, SDK and API changes, open-source releases, engineering write-ups, and Show HN projects.
Revolut reduced AI coding inference costs ~90% by replacing Claude Code with self-hosted GLM models, enabling broader internal developer adoption.
WorkOS data breach involved attacker using ARTEX AI to compromise internal sales app; threat actor accessed customer contact records and communication summaries.
Anthropic open-sources FlashProve, cryptographic verification ensuring API responses originate from specified models, preventing router substitution.
Qwen 3.8 27B running locally can also interpret and read model weights, suggesting the capability is not unique to Astra.
Weight Space Programming enables LLMs to directly edit neural network weights; demonstrated on Qwen and DeepSeek for reasoning optimization.
Trajectory raises $45M Series A from Sequoia for non-technical AI model post-training customization tools.
llm 0.37 adds GPT-6.1 variants, OpenAI-compatible endpoint config, defaults to cheaper GPT-6-Luna.
OpenAI publishes internal engineering practices for large-scale Git repository management and version control.
Trajectory raises $45M Series B at $300M valuation to democratize AI model training for non-technical users.
Legora CTO discusses AI evaluation techniques, tabular data handling, and billing model disruption for legal AI applications.
