Newsroom

News

Latest news and insights from the Curtail team.

September 25, 2026

The Slowest Part of an Agent Loop Is the Agent Working Blind

Productivity rose 209% with a behavioral diff in the loop, on 68 of 71 models. The run barely sped up; the agent stopped doing the same work twice.

ReGrade · AI Coding · Developer Productivity

Read article→

September 25, 2026

What AI Coding Agents Break When Nobody Is Looking

We measured unintended changes to working code in 74 models from 23 labs. Hallucinations fell over 79% on the typical model, and on all 73 we could score.

ReGrade · AI Coding · Hallucinations

Read article→

September 25, 2026

Your Agent's Bill Is Not the Problem. Your Cost Per Fix Is.

A behavioral diff makes each agent run spend about 30% more tokens. It makes each bug fixed cost 58% less. Only one of those is what you buy.

ReGrade · AI Coding · Cost

Read article→

June 26, 2026

Claude Mythos Is Now Hunting Your Memory-Safety Bugs. Migrate to a Memory-Safe Language, and Prove the Rewrite Behaves.

Claude Mythos found a 16-year-old memory-corruption bug in FFmpeg. Migrate memory-unsafe C/C++ to a memory-safe language, and prove the rewrite behaves.

Memory Safety · Rust · Refactoring

Read article→

June 16, 2026

The Most Expensive Thing Your AI Does Is Re-Read and Rework Its Own Code

AI coding agents bill by the token, and much of that bill pays for the agent to re-derive what just changed. Here is that line item and what controls it.

AI Code · Engineering Leadership · Cost Management

Read article→

June 10, 2026

How to Catch AI-Generated Code Regressions Before They Ship

AI-generated code passes tests and still changes behavior the tests don't assert. How to catch the runtime regressions your CI lets through.

AI Code · Regression Testing · CI

Read article→

April 3, 2026

ReGrade 3: The Self-Healing Loop for AI-Generated Code

AI coding tools produce 1.7× more bugs and 2.74× more security flaws than human code. ReGrade 3 gives your coding agent a behavioral feedback loop.

ReGrade · AI Coding · MCP

Read article→

April 3, 2026

ReGrade 3: Refactor With Confidence, AI-Powered Migration Without Behavioral Risk

AI can speed up a rewrite to a memory-safe language. ReGrade 3 checks, field by field, that the refactored code behaves the same as the original.

ReGrade · Technical Debt · Memory Safety

Read article→

April 3, 2026

ReGrade 3: Every Merge Request Gets a Behavioral Audit, Automatically

ReGrade 3 runs in your GitLab or GitHub pipeline and posts field-level behavioral regression reports in your merge request comments.

ReGrade · CI/CD · Merge Requests

Read article→

April 3, 2026

ReGrade 3: Deterministic Guardrails for AI-Generated Code

AI coding tools ship code faster, and bugs with it. ReGrade 3 compares behavior deterministically before, during and after the merge request.

ReGrade · AI Coding · Security

Read article→

April 2, 2026

How We Found a 7-Year-Old Vulnerability on the First Replay

A widely used collaboration platform shipped a password hash disclosure bug for 7 years. ReGrade caught it on the first replay, without knowing it existed.

ReGrade · Security · Case Study

Read article→

February 23, 2026

Playwright MCP Gave AI Eyes for the Browser. ReGrade Does the Same for Your Network.

Playwright MCP gave AI agents structured browser state. ReGrade gives them field-level network behavior and flags API regressions and missing headers.

MCP · API Testing · ReGrade

Read article→

February 6, 2026

The Hidden Risks of Agentic AI Systems, and Why Verification Matters

Agentic AI systems that let LLMs fetch and analyze data fail in documented ways, from silent hallucinations to context window degradation.

Agentic AI · LLM Risks · Verification

Read article→

January 20, 2026

LLMs: Strengths and Weaknesses

We treat LLMs like reasoning machines and act surprised when they get math, logic or proofs wrong. CTO Robert Ross on the limits behind those failures.

AI · LLM · Machine Learning

Read article→

November 24, 2025