Two years of agentic coding: what stuck
The workflows that survived contact with real work, the ones that did not, and what the whole thing actually changed.
tech, developers, and the code underneath
22 pieces tagged coding-agents,
from January 31, 2025 to July 31, 2026.
The workflows that survived contact with real work, the ones that did not, and what the whole thing actually changed.
Large mechanical refactors are the clearest win available from coding agents. Here is the process that keeps them safe.
Running many coding agents at once works better than expected on one axis and worse on every other. What actually happens.
The analogy is exact except for the part that matters: there is no parameterized query for natural language.
A convention nobody standardized became standard anyway. Here's what belongs in it and what doesn't.
A frontier release with a large price cut, plus effort controls and context compaction as a first-class feature.
Google ships a frontier model and Antigravity, an agent-first development environment. The bundling is the strategy.
OpenAI ships a Chromium-based browser with an agent that can act on pages. The prompt injection surface is now your whole session.
OpenAI ships an agent builder, an app SDK, and a distribution channel. The strategy is now unambiguous.
A model tuned for long-horizon autonomous work, plus checkpoints and context editing in the SDK.
OpenAI merges Operator and deep research into one agent with a virtual computer. The interesting part is the permission model.
An acquisition collapses, Google licenses the technology and hires the founders, Cognition buys the rest. Everyone learns something about acquihires.