first big launch at the new job 🚀... of course it was going to be developer education.

Come see me present about the future of AGI at the CodeRabbit hackathon today!

Wrote a short blog post to articulate my takeaways from conversations with frontier labs. By sitting between models and users working on real world tasks, coding agent platforms have become key players in accelerating progress toward AGI. They have tremendous LEVERAGE and

Cline is an AI skill training company. We build the skill tree model labs need to achieve AGI.

When we say we stick to staying open, we mean it. Welcome Cline

Turns you don’t need a presentation layer to train agentic skills to models. Agent experience doesn’t require an engineer seeing what’s going on. The SDLC of the future will be about slowly reducing the human in the loop without sacrificing quality until we achieve AGI

Imagine using an agent framework in 2026… couldn’t be Cline

Leaving Cline has consequences Lan

The Cline mafia doesn’t allow people to leave

The anxiety I get when my context window is above 50%. I cannot wait for a real 1m token coding model like Gemini 3 to come out next week

Whenever I go and try cursor or Claude Code, I’m reminded by the design decisions that make Cline awesome. We might not have as big of a budget, but we make the engineer using Cline into a special forces engineer who can build anything in a fraction of the time

Very excited to be launching a new stealth model. The era where Google is the only 1 million token model is over. With more context comes: - more research - longer planning cycles - going straight to act without compacting - executing on more complex tasks

I wake up in the middle of the night, in a cold sweat, terrified, from a vivid nightmare where Ryo Lu violently slits my throat and leaves me to bleed out on the sidewalk. I wake up and am relieved that it was just a dream.

Everyone is talking about SF lacking events but my B2B SaaS just added OpenTelemetry and now we are producing millions of events

Vibe coding: when the AI cracks the tough problem, but you're left sipping coffee, questioning if you're even a coder anymore.

Supporting multiple models is challenging. New models emerge weekly, often requiring significant effort to support fully. - We write new prompts and warnings - Develop new tooling calling systems - Create new evaluations for error states Our team barely sleeps.

I love open-source, but you're kidding yourself if you think open-source models are nearly as good as frontier models at agentic coding.

Imagine missing Cline's seed round just to fork it and add a paywall.