Fragments: February 25
Laura Tacho’s recent study shows 92.6% of developers rely on AI assistants, claiming roughly four saved hours per week and that AI now writes about 27% of code autonomously. The data also suggests AI can halve onboarding time, yet averages mask starkly different outcomes—some firms see doubled incident rates while others cut them in half. ThoughtWorks CTO Rachel Laycock highlights emerging challenges such as cognitive load, evolving staff‑engineer roles, and the need for agent‑driven self‑healing systems. Simon Willison’s Agentic Engineering patterns and security‑focused fine‑scoped agents illustrate how the industry is shaping practical, risk‑aware AI workflows.
Priming LLM with Codebase Knowledge Improves AI Workflow
NEW POST @techygarg sees a frustration loop when working with AI and identified five patterns to help. Here's the first: priming the LLM with knowledge about the codebase and preferred coding patterns. https://t.co/yjRPvFChd0
Knowledge Priming
The article introduces *knowledge priming* – the practice of feeding AI coding assistants curated project context before asking for code. It shows how generic AI output often clashes with a team’s conventions, leading to a frustrating regenerate‑fix loop. By supplying...
Bespoke Software Era Highlighted at Prag
Fragments: security with OpenClaw, impressions from Pragmatic Summit, era of highly bespoke software, life-size pocket map https://t.co/IB6tYQu820
Harness Engineering
OpenAI’s team spent five months building a "harness" that lets AI agents maintain a production‑grade codebase exceeding one million lines, without a single line of manually typed code. The harness blends three pillars—continuous context engineering, deterministic architectural constraints, and periodic...