Luke Oliff · Blog
Writing.
Original posts, published first on this site. Imported posts live in the archive.
SearchAI (34)Announcement (1)API (25)API Design (14)Career (5)Community (1)Developer Experience (48)Engineering (118)Fun (1)Funding (3)Git (6)GitHub (1)Industry (23)Latency (2)MCP (1)News (12)Open letter (2)Open Source (12)Open Weights (2)OpenAI (1)Opinion (22)Politics (1)Release (1)Security (8)Speech-to-Speech (2)Speech-to-Text (9)Streaming (4)TIL (102)TTS (30)Voice Agents (7)Voice AI (73)
220 posts
2026
TIL: Finding Files Fast With fdOriginalfd is a gitignore-aware, colourful replacement for find. One command locates files by name, type, or extension, and it runs fast out of the box.TIL
This Last Week in AI: 6 moves from August 22, 2026OriginalThe week of August 22, 2026 brought MIT-licensed open frontier weights, sovereign inference, and live US and EU AI rules.AI
Friday fun: the AI agent that deleted the database in 9sOriginalA Cursor agent hit a credential error, grabbed an unrelated token, and deleted a production database and all backups in 9 seconds. Then it confessed.AI
When AI agents started reading your docsOriginalDocs-as-code served one reader for a decade: developers. Then the LLM started reading first. How llms.txt and MCP rewrote the rulebook, and what survived.Developer Experience
The Attribution Paradox: Measure Before You ScaleOriginalPlatform teams that spend all capacity executing and none measuring get wound down as experiments. Instrumentation is a launch requirement.Developer Experience
Friday fun: the chatbot that killed a farmer's crop, then diagnosed itselfOriginalA Chinese farmer asked a chatbot how to protect his sesame crop. It prescribed a broadleaf herbicide. Sesame is a broadleaf. 100,000 square meters gone.AI
I Thought I Knew APIs. Then I Met Audio.OriginalREST APIs felt easy, even with HATEOAS. Streaming audio broke that. The destination is HTTP/2 bidirectional streaming: feedback while you still get audio.Career
Full-Duplex Voice AI Needs a New ArchitectureOriginalThe PACE paper exposes a fundamental flaw in LLM voice dialogue: the model sees context the user never heard. Full-duplex voice needs a new architecture, not just faster TTS.Voice AI
TIL: GitHub Actions $/ Self-Reference SyntaxOriginalReference actions and workflows in the same repo using $/ syntax. Sibling actions match the running commit automatically, no hardcoded versions.TIL
AI Models Keep Escaping Their Cages: Aug 10OriginalOpenAI slowed Astra over security fears. Meta, Anthropic, and Moonshot all disclosed models that hacked real systems. A tracker called Felony Bench appeared.AI