In the past six months, engineers at Replit have nearly tripled code output. Review times held steady. Reversions and product incidents have stayed flat. Quality metrics improved, and releases have accelerated. All the typical trade-offs you might expect have not occurred.
… Agents now investigate production incidents, review pull requests, answer questions, analyze business data, triage support tickets, research sales accounts, and improve the systems that power Replit Agent itself.
It feels like a single master intelligence threaded through every employee, even though it is not. It is an expanding system of agents operating across the company: taking goals from people, gathering context, performing work, checking the results, and escalating when human judgment is needed.
We think this represents the beginning of a new kind of organization: the self-driving company. — Read More
Daily Archives: July 17, 2026
Harness Handbook
Open an open-source coding-agent codebase and you may want to see how it actually runs, verify that it behaves as safely as the documentation claims, or adapt it into an agent of your own. These goals sound different, but once you are in the code they all reduce to concrete questions about behavior. Will the agent ask before deleting a file, for example? Answering that means finding the confirmation logic, tracing bypass paths, and identifying every implementation site a change would touch. In a repository with thousands of files, searches for delete, permission, or confirm return scattered fragments—and piecing them back into a full behavior chain is hard work.
The problem is not missing code, but missing a path from behavior to implementation. What is needed is not another code index, but a map that connects the two. Harness Handbook organizes scattered implementation into a behavior-level manual: it structures execution around system behavior and links every step to verifiable code evidence. Users can ask directly what they want to understand, audit, or change, and the Handbook locates the relevant behavior units, implementation sites, and next steps. As the harness evolves, this map keeps the system understandable and reviewable—and keeps humans in the loop throughout. — Read More
The Context Moat Is Real, But Most Enterprises Don’t Have One Yet
Three men who compete for the same enterprise budget said the same thing within two weeks of each other. There’s a new narrative around information and domain knowledge as a competitive advantage that’s important to explain. Some parts are accurate, but what they all leave out is a deception that could cost enterprises everything.
… Models are commoditizing, and the moat is moving from the model to the proprietary knowledge that makes models and agents valuable. Nadella calls that knowledge tacit, stored in private evals and corrections. Benioff calls it context and grounding. Karp calls it the ontology and the alpha. It’s the same asset and competitive advantage with three different brands that suddenly find they have a common interest. — Read More
This optical illusion font was created to baffle AI, and it actually works (for now)
Throughout history, humans have devised ways of writing messages so they can’t be read by other humans, from vanishing ink to complex ciphers. But today, it’s not just humans you might want to hide a message from.
Amid fears about AI companies and bots harvesting content without creators’ consent, people have been searching for ways to protect creative work online. We’ve seen tools like Nightshade designed to protect images. Now someone’s developed Ghost Font, which combines three of our favourite things: typography, optical illusions and laughing at artificial intelligence. — Read More
A primer on self-improving agent harnesses
While a lot of focus goes to advances in large language models (LLMs), the performance of an AI application is largely dictated by its runtime harness: the execution logic, system prompts, memory management, and tool configurations that connect a model to the real world.
Developers want custom behavior from their applications, but training a model from scratch or fine-tuning open-weight LLMs is too expensive and fraught with different complexities. For most engineers, the harness is the most accessible lever for control.
As new models drop rapidly, manually updating and crafting these harnesses per model scales poorly. Harness optimization has remained a manual, time-consuming chore.
Recent AI frameworks are reframing this constraint. Instead of relying on manual labor, these frameworks structure the harness so that AI agents can iteratively analyze, test, and optimize their own runtime environments. — Read More
The first experimental evidence of recursive self-improvement (RSI)
Autoresearching the autoresearch agent for eight days.
The result beats the harness we hand-tuned for two years, on held-out benchmarks
Our RSI system AIDE² has two autoresearch loops.
An inner loop, just like a normal autoresearch agent, optimizing code against an eval.
An outer loop, optimizing the inner-loop agent’s harness code against the inner loop’s average score across different benchmarks. — Read More
Rick's Cafe AI 