LLMs

Fabricated case citations generated by AI are appearing in court filings at an accelerating rate. Combined with other tracking efforts, we have identified over 1,000 filings containing hallucinated citations from self-represented (pro se) litigants and lawyers alike.

Watch now | The AI work your company cannot see is the AI work your company cannot learn from.

The production trade-offs that only appear once your model is live.

These MCP servers make my local LLM even better.

For AI engineers who want to understand every step, not just call the library

Zaya1-8B is a huge shift in LLMs, and the results are impressive.

Most LLM failures in production aren’t random — they’re predictable. I kept hitting broken JSON, silent failures, and outages that froze my entire app. Prompt engineering didn’t fix it. So I built a control layer above the model — and took structured output reliability from 0% to 100% without changing a single prompt.

Think of it as the Linux desktop problem, all over again

A 1,000 libraries, 90% coverage, 1,700 in API tokens. Nobody typed a single test by hand.

Everybody is building agent skills, but not all skills are created equal. Here are some recent research papers that empirically show best practices to build them.

A coding agent is the model plus everything you build around it. Harness engineering treats that scaffolding as a real artifact, and it tightens every time the agent slips.

The bet every serious developer needs to make on on their agent stack

Revisiting the PocketOS Incident

Companies everywhere are trying to leverage AI to boost internal productivity metrics. Some, like Ramp and Intercom, are succeeding. Many are failing.To

As a pandas core contributor and early Parquet adopter who built AI data pipelines at streaming company Tubi TV, Chang She saw firsthand why the traditional data stack breaks down for AI workloads—and founded LanceDB to fix it. Chang joined Ben Lorica to explain why vector databases are too narrow a solution for modern AI …

A conversation about running a startup alone, with agents doing the work of a full engineering team

I just sat in a room full of data engineers the other week who were worrying about AI automating them out of work the same way auto manufacturing in Detroit

Rethinking personal AI and agents

An Argument into the Agora

AI-coding tools don't create bad engineering practices – they accelerate them. Here’s how to fix your delivery process before the next midnight page.

Been running local models as part of my daily workflow for a while now, and what surprised me most is how often local turned out to be the better choice, not a compromise.

Anthropic has opened its Claude Hardware Interface (Bluetooth API) to developers, enabling an ESP32-S3-based desk companion to connect directly to the

Manage a team of AI agents to run your business. Org charts, budgets, governance, and goals — all in one deployment.

The concept of orchestrating agents is gradually spreading among developers. Claude has planned, or you may already be able to use (depending on when

The deck is a bit stacked, folks.

Swamp Club — your agent builds typed models, wires workflows, and stores versioned results. AI-native automation that extends itself.

Nearly once a week I receive an email from a different stranger. The messages are eerily similar. The sender has developed an unusual relationship with an AI gained over many hours of interactions. The AI has given them extraordinary insight … Continue reading →

Watch now | The next serious agent failure won’t look like a jailbreak.

Back in 2024 I wrote that AI helps me remove boring work at the margins. This is fine for a lone writer, but how to scale this to an entire team of technical writers? How to make the system helpful but not intrusive? These are all questions I’m starting to answer now, partly through experimentation, but also through dialogue with practitioners and colleagues. One answer I’m testing these days relies on GitHub Agentic Workflows.