Latest News

Catch by Recovery Policy, Not Convenience

Liquid syntax error: 'raw' tag was never closed

I Thought My Multi-Agent Debate Engine Was Broken. The Real Bug Was the Prompt.

v0.2.1 RELEASED — Aug 28, 2026. Release notes · Field test report · PyPI v0.2.1 Update: The 2,333->359 join collapse described in this article is…

I Asked a Free Model the Same Question for 48 Hours. The Drift Was the Signal.

Most model benchmarks tell you how smart the model is on the first attempt, which is almost never the problem in production. The real problem is wha…

The Golden-File Refactor Loop: Record, Verify, Move, Commit

You do not understand the messy function. That is fine. The snapshot does not care about your understanding. Golden files turn "I think this is safe…

The Gate Said No. Now What? A Triage Procedure for Rejected Agent Patches

A gate that rejects a patch is only half a policy. The other half is what happens after the rejection. In most pipelines, a failing agent patch prod…

A static analysis rule written from a spec is a hypothesis

Last week I shipped a tool that predicts which ComfyUI custom nodes will break on your next git pull . It answered one question: does every name a p…

How I find out whether a filter is filtering: ask it two questions that cannot share an answer

A tool that filters can fail in a way that is almost impossible to notice: it returns a plausible number of plausible results, and none of them were…

How AI Assists in Cross-platform Development (2026 Data)

Originally published at nlocoding.com 57% of cross-platform apps miss revenue targets due to poor platform optimization. (Source: Forrester, 2026) W…

Most AI Second Opinions Are Theater. I Built a System That Actually Fights Back.

The first time I ran two LLMs against the same pull request, 89% of their "debate" was fake. Not wrong. Not low-quality. Fake. The second model was…

No, the LLM Doesn't Get to Approve Your Refund

ADR 001: why refund eligibility is deterministic Java, not a model judgment Part 3 of an ongoing experiment: building an LLM-powered support agent w…

Change-Detector Tests Considered Harmful (2015)

The linter ecosystem is actively migrating silent failure into an undetectable idiom

In a well-known family of LLM evaluation bugs, the failure doesn't look like a failure. A judge model goes down, and the benchmark keeps publishing…

Making HTTP Fail on Purpose: Building a Small Chaos Library for Java - Flaky HTTP

I recently built and open-sourced Flaky HTTP , a small Java 11 library for deliberately making HTTP calls less reliable. That may sound like an unus…

A test said the server started. I deleted the server. It still passed.

Here is a test from a real, well run Node project: test ( ' server starts ' , async ( t ) => { const app = build () await app . listen ({ port :…

Your Free AI Server Has a Ceiling. Measure It in 30 Minutes Before the Team Does

Tuesday, 10:47 AM. Fourteen developers open their IDE extensions at once, and the shared AI server starts returning timeouts. Nobody planned for the…

I Built a Company That Doesn't Exist to Test an AI Product

These articles come from lessons learned while building Eterna Clarity and the operating system I use to run it. One of the hardest test environment…

I Got 24/24. I Still Didn't Open the Final Test.

These articles come from lessons learned while building Eterna Clarity and the operating system I use to run it. A local model hit 24 out of 24 on t…

Claude Structured Outputs Refusal Handling: Stop Parsing HTTP 200 Refusals

Claude structured outputs refusal handling belongs before domain deserialization. A successful HTTP exchange only says the API accepted and processe…

My Agent Refused 96 Times. That Was the Right Output.

In the last article , I wrote about a release story that was weaker than the engine underneath it. This time the mistake was worse: I was rewarding…

Your Keep-Alive Is Lying to You: Six Connection Myths I Measured on a Free Model Server

Every code review has the same advice. "Add a retry." "Set a timeout." "Use keep-alive." Nobody measures first. I got tired of guessing. So I built…

.NET 10 ActivitySamplingResult PropagationData: Why Recorded Turns False

.NET 10 changed a small tracing rule that can quietly invalidate a custom sampler. With .NET 10 ActivitySamplingResult PropagationData , a child act…

Testing Production AI Agents: A Practical Framework for Graph-Based Agent Systems

Table of Contents Introduction Why Testing AI Agents Is Different The Anatomy of a Production AI Agent Why Traditional Unit Testing Becomes Difficul…

Agent Patches Get Three Gates: Property Checks, Pinned Fixtures, and a Flaky Freeze

An agent's test suite is green. That is the least informative sentence in modern software engineering. Example tests encode what you already thought…

Vizra Evals and Pest's Evals Plugin: When You Want Which

If you are testing AI agents in Laravel, there are now two packages with "evals" in the description, and the obvious question is whether you need bo…

The model said it read the report. It didn't.

I'm a regular automation engineer. Over a few months I built five small projects that test local AI models, and one problem kept coming back: the mo…

Zero Is Not an Empty Value in Financial Software

What does zero downside mean in a cloud commitment recommendation? It should mean that the modeled draws produced no loss after commitment cost, unu…

Organizing Application-Specific Artifacts in Automation Frameworks

Maintaining a clean and modular repository structure is essential when scaling test automation frameworks across multiple applications. A core pract…

Why API Testing is the Silent Backbone of Great Software

Most of what makes a modern application work never appears on screen. Behind every screen tap or page load sits a network of APIs quietly passing da…

Doc Drift Is a Bug: A Claim-Checking Loop for Model-Generated Docs

Generated docs fail most often at the boundary between prose and code, where a flag name or a signature silently disagrees with the source. A free-t…

AI Is Transforming Software Development Workflows

The highest-value use of Claude Code is rarely the one in the demos, and it isn't writing code. It's answering a question like "where does the webho…