Install
Please confirm you are human
This browser or connection looks automated. Press and continuously hold the control for 3 seconds to enable Google-hosted web results and, when separately allowed, AI-assisted answers.
A successful check enables 100 search requests. Interactive access does not authorize scraping, systematic collection, or reuse of search output.
News
An HTTP 200 is not a delivered job application
10+ hour ago (336+ words) Every auto-apply tool I have taken apart marks an application "sent" at the same moment: the instant it dispatches the click. That is a read-back of our own input, and it is the weakest signal in the whole pipeline. I…...
Why a PR walkthrough can pass validation and still be hard to read
12+ hour, 35+ min ago (770+ words) I made an HTML guide for reading pull requests. When I shared it with coworkers, the visual feedback was encouraging, but I still felt that parts were hard to follow. One person wanted to see the explanation before the code…...
What Happens When You Accept Website Cookies?
8+ hour, 9+ min ago (443+ words) An easy guide to browser cookies, privacy, and online tracking When you click “Accept cookies,” a website may ask your browser to store small pieces …...
Can a Local Coding Model Save You Money?
18+ hour, 17+ min ago (154+ words) The practical question is whether avoiding a hosted inference charge still looks economical after the machine, electricity, maintenance, review time, and fallback calls enter the calculation. The measurements cover task outcomes, elapsed time, and API charges. The worksheet leaves equipment,…...
Ollama's Responses API accepts previous_response_id, returns 200, and forgets the whole conversation
15+ hour, 32+ min ago (538+ words) Ollama 0.34.0, Debian 13, CPU only, qwen2.5:1.5b. The Responses API, which is the OpenAI-compatible endpoint Codex and similar clients use. Turn 1 plants something the model could not guess: password. Now two controls. The same question with the full history sent in the request…...
My Harness Used One Label for Three Different Failures.
15+ hour, 25+ min ago (1083+ words) Three fixtures, three separate calls into the same reducer. Here is the complete failure_reasons each one returned, unedited: These are minimal fixtures with no sandbox event and no tool response, so the first and last codes fire in all three and…...
Playwright email verification can pass for the wrong reason: six checks for a trustworthy test
16+ hour, 19+ min ago (575+ words) We deliberately made an email-verification test finish on the wrong account. The page said Email verified. Playwright passed. Then we added an assertion for which account had been verified. The same scenario failed. This was a controlled browser experiment with…...
AI Model Evaluation: Best Practices for Testing and Validation
17+ hour, 11+ min ago (19+ words) Evaluating AI models is... Tagged with ai, evaluation, machinelearning, testing....
Two "Codex CLI" models on the same benchmark: the harness hides the model
17+ hour, 25+ min ago (260+ words) Specific Labs dropped Real-SWE, an enterprise-code SWE benchmark, and the leaderboard is a great study in why you should never read "Claude Code" or "Codex CLI" as a model name. Same harness, two different brains: Same vendor's CLI, same harness,…...
My Scanner Passed Until I Built a Harness That Lied to It on Purpose
17+ hour, 39+ min ago (271+ words) Every security tool has a dirty secret: nobody tests whether the detector detects. The tests pass, the CI is green, and everyone moves on. Today I pointed my own scanner, mcpscan, at a set of synthetic attack files I wrote…...