Most important take away

Title and promotion structures inside large engineering organizations quietly redirect engineers away from building the product and toward gaming a rubric — the moment Confluent introduced titles to a previously flat 500-engineer org, output shifted from product work to promotion politics overnight. To do your best work, pick a company whose mission you care about, optimize for product quality over career ladder, and lay strong technical groundwork up front rather than fighting fires later.

Summary

Career advice

• A CS degree is more useful in 2026 than it was in the 90s; curricula have caught up to industry languages, so finish school quickly rather than dropping out.
• For European engineers, the single most important career move is getting to Silicon Valley for 3–4 years. The fastest path: join a US-based company while still in Europe, then transfer over via a manager track.
• Consulting is a career trap — Thomas calls his nine years of consulting in Finland “the biggest regret of my life” because you ship and forget rather than building something you own.
• Only start a company if you feel there is literally no other option; conviction is required because the default outcome is failure and burnout.
• Don’t take a manager role just because you were asked. Thomas accepted a frontline manager role at Uber out of people-pleasing and spent three years shielding his team from politics instead of building product.
• If you’re told to “care less” or “tone it down” to get promoted, that’s a signal to leave.
• To stand out: pick a company where you genuinely want to work on the product, then optimize for the product rather than the rubric.

Org design / promotion culture

• At Linear, >90% of engineers carry the title “staff engineer.” There are effectively no levels, though compensation does vary with contribution.
• Titles are introduced in large orgs because frontline managers’ job-to-be-done is promoting their reports — without titles, that role loses purpose.
• The Confluent anecdote: a flat 500-engineer org worked well until managers lobbied for titles; the day titles shipped, engineers stopped optimizing for product and started arguing about their level.
• Linear keeps the manager layer thin — a 65-person product team has only ~3 managers, and even the CEO still designs and the CTO still codes 80% of his time.

Product / engineering patterns at Linear

• One-word strategy: quality. In a crowded category (project management has hundreds of competitors), the only way to pull users off entrenched tools is to be ~10x better, because switching cost is high.
• Zero-bug policy. Bugs accumulate at a constant rate as features ship, so you have to fix them eventually anyway. Paying down the backlog once (took Linear ~3–4 weeks) and then fixing every new bug immediately costs the same total work but yields: (1) a bug-free product, (2) engineers who introduce fewer bugs because they know they’ll have to fix them now, (3) delighted customers who get fixes within ~12 hours.
• Lay groundwork before product. Linear spent ~6 months building a sync engine and stood up a Kubernetes cluster on GCP on day one — things a normal pre-PMF startup would never do. The payoff: they see infrastructure bottlenecks 1–2 years in advance and never fight fires (vs. Uber’s pattern of crappy infra → 2 years of firefighting → burned-out teams → 4-year rebuild).
• Speak up. Thomas’s view is that the politeness/politics culture in US big-co engineering suppresses the truth-telling that actually improves products.

AI / agents

• Linear treats agents as team members. Triage (the inbox for inbound issues from Sentry, Intercom, support, etc.) can be automated: an agent triages, labels, assigns, and a coding agent attempts a PR. Current success rate is ~30%; expected to climb to 50–60% with newer models (Opus 4.5, Codex Feat. 3).
• Linear’s role is shifting from issue tracking toward project management: helping humans decide what to build and orchestrating fleets of coding agents that handle the how.
• Bullish on product engineers (good taste + AI = huge leverage). Bearish on pure pipe-moving engineers — those jobs get absorbed by AI.
• Limits of AI today: it cannot generate genuinely novel design (only remix existing patterns), and it cannot feel a 150ms vs 200ms animation as a user does. Without that emotional signal, it can’t drive UX quality.
• Cost: agent compute will likely cost more than a human engineer at first as orgs scale to thousands of agents per problem, then trend toward zero over the long term — though the next decade is uncertain given GPU/power/data-center shortages.

Chapter Summaries

Education and early career (Finland, the 90s). Thomas dropped out of University of Helsinki after one week in 1996 because they were teaching Pascal while industry had moved to HTML and Java. He’d advise his 18-year-old self in 2026 to actually finish school — curricula have caught up.

The consulting trap. Nine years running his own consulting shop in Finland was lucrative but, in retrospect, wasted time. Real growth required getting to Silicon Valley, which happened almost by accident via a friend at Groupon.

Uber: the manager detour. Thomas accepted a frontline manager role out of obligation, spent three years on politics he hated, and returned to IC work as a senior staff engineer. The lesson: don’t accept management just because you’re asked.

The promotion game. At small companies, engineers optimize for product. At ~1,000+ engineers, they optimize for the rubric — building impressive-looking projects that don’t help the company. Thomas left Uber partly because his manager told him to “care less” and tone down honest feedback.

No-titles experiment (Confluent story). A flat 500-engineer org worked well until frontline managers lobbied for titles. The day titles shipped, focus shifted from product to “why am I only a senior?” Linear is trying to keep titles flat as long as possible.

Why a CTO still codes. Thomas spends ~80% of his time coding because he intentionally hired managers and infra leads to absorb his other responsibilities. Linear’s whole org is structured to keep IC work primary — 65-person product team, only ~3 managers.

Linear’s one-word strategy: quality. In a category with hundreds of competitors, only a ~10x better product justifies the switching cost. You can’t retrofit quality culture onto a large team — you have to hire for it from day one.

Zero-bug policy. Bugs come in at a constant rate, so the total fix-work is the same whether you let them pile up or fix immediately. Fixing immediately yields a clean product, fewer new bugs, and delighted customers.

Laying groundwork. Six months on a sync engine and Kubernetes on day one let Linear see bottlenecks 1–2 years out and avoid the firefighting cycle Thomas saw at Uber.

Agents in Linear. Triage automation routes inbound issues to coding agents that attempt a PR; ~30% success today. Linear is evolving from issue tracking toward orchestration of agent fleets and helping PMs synthesize inbound context.

Bullish/bearish on software careers. Product engineers with taste are massively leveraged. Pure data-pipe engineers are at risk. Designers and PMs still matter because AI can’t yet generate novel design or feel UX latency.

The future of AI cost and capability. Three years ago AI couldn’t write a line of code; now it implements features. Thomas keeps catching himself moving goalposts. Long-term he expects compute cost to trend toward zero, with significant turbulence (GPU shortages, depreciation risk) in between. Hardest remaining problem: instilling human emotional judgment into AI.