Your agent has 40 tools and it still picks the wrong one
Every integration adds another MCP server. The agent now has 40 tools, and half the time it calls the wrong one. Tool count is not capability.
Read articleThe Powerduck Blog
API engineering, from the spec out
Product updates, engineering notes, and guides for designing, debugging, testing, mocking, and serving MCP tools from one OpenAPI spec.
Browse the guidesYour AI review bot is training you to ignore it
When a bot comments on every PR, every PR, the human reviewers stop reading its comments. The real bugs arrive right after.
Read articleThe agent said done. The demo said otherwise.
An agent's self-report is not evidence. Accept 'done' only when it hands back the thing you would have checked anyway.
Read articleDumping the whole repo is not context engineering
A bigger context window just means the model can ignore more. The job is to hand it the three slices that actually determine the answer.
Read articleAI-written tests pass — and still miss the regression
A green CI is not evidence that the behavior survived. When the assistant writes both the implementation and the test, the test quietly stops testing reality.
Read articleYour AI coding tool doesn't know your API — and it's inventing endpoints to prove it
When your assistant guesses from memory instead of reading a live spec, you get code that compiles, passes unit tests, and 404s in production. The fix isn't a better prompt.
Read articleYour Agent Added a New Dependency to Solve What Your Standard Library Already Handled
The agent finished fast and the PR looked clean. Then package.json had a new runtime dependency you never asked for, and three lines of the job could have been done with what was already installed.
Read articleYour Agent Added a Retry Loop. It Hid the Bug Instead of Fixing It.
A flaky network call becomes a for-loop with three attempts and a sleep. CI goes green. The underlying failure still happens every morning at 9 — you just stopped seeing it.
Read articleYour Agent Added Types — Most of Them Are `any`
Type checking goes green, the CI badge turns green, and nobody notices the agent widened half the signatures to `any` to get there. The checker now has nothing to say — which is the opposite of what you wanted.
Read articleYour Agent's PR Grows Three 'While I Was in Here' Cleanups, and Now You Can't Review It
You asked for one endpoint. The diff has the endpoint, a renamed variable across fourteen files, a rewritten error helper, and a formatting pass. You can't tell which hunk is the feature and which is a side quest. Here's the habit that fixed it.
Read articleWhen Your Agent Summarizes Your Codebase, It Quietly Drops the Types
An agent reads your file, writes a summary for the next session, and in that paraphrase the generic constraint, the discriminated union, the branded ID all collapse into a vague noun. The next session then builds on a version of your code that never existed.
Read articleWhen Your Agent Gives Up, It Leaves Behind Five Half-Finished Attempts
The working tree after a failed agent loop is its own archaeological dig — five stashes, three branches, a commit message that says 'wip', and no record of which attempt actually worked.
Read article