Jori Lallo on Linear Agents, AI coding tools, and why they don't chase AI hype

Summary

Jori Lallo, co-founder of Linear, makes a deliberate case for why the company has stayed quiet during the AI hype cycle — and why that is starting to change.

Linear builds project management and issue-tracking software, used heavily by engineering teams. Its brand is built around speed, craft, and restraint: chasing milliseconds on load times, shipping purposefully rather than loudly. That philosophy created an early friction with AI tooling. LLMs carry inherent latency, and until recently, Lallo says, that felt fundamentally incompatible with what Linear is trying to be.

The turning point was DeepSeek's thinking mode. Lallo describes it as a light-bulb moment — not because the model was transformative on its own, but because it shifted user expectations. People now anticipate a pause before a high-quality answer. That new UI paradigm gave Linear a way to integrate AI without compromising the product feel. The company has been in course-correction mode for roughly six months.

Linear Agents

Linear launched agents this week, positioned as the interface layer between engineering teams and whatever coding agents they choose to run. Lallo is deliberately agnostic on which agents win — Claude, Gemini, OpenAI — describing Linear as the workflow surface rather than a participant in that race. Today, agents inside Linear behave like regular users, posting progress updates as comments in issue threads. Lallo acknowledges that's a hack, and the team is still figuring out what a purpose-built agent experience actually looks like inside the product.

The near-term roadmap points toward non-coding agents running alongside coding agents. Lallo's example: a feature-flagging agent, a coding agent, and a marketing agent collaborating on a single task — say, writing the changelog blog post after a feature ships.

On open standards

Lallo is broadly supportive of MCP and open APIs, framing it as a welcome reversal of the past decade's platform lock-in. His practical caveat is security: Linear holds a lot of customer data, and openness has limits when it comes to what the company actively promotes versus what it technically permits. His read on computer-use agents — tools that can puppeteer a mouse and keyboard — is that they're the ultimate band-aid, capable of doing anything but not purpose-built for anything.

The AI adoption posture

Linear doesn't mandate tools internally. Engineers try what interests them, and adoption is organic. Lallo draws a contrast with companies that announce they've replaced staff with AI — naming Klarna without naming them — and treats that as a legitimate marketing strategy that simply isn't Linear's. The company started experimenting with GPT-era tools when everyone else did, hit the familiar 70% quality ceiling, and pulled back rather than ship something unreliable. That restraint meant Linear wasn't first to market with AI features, but Lallo argues customers buying product rather than buying hype will still be there when the product catches up.

The commercial signal is that demand for AI tooling is now strong enough that Linear is investing heavily. The pitch to those customers is straightforward: Linear was already where engineering teams do their work, so agents integrating there is the path of least resistance.