Field notes from inside frontier hiring.
Comp data we wish someone had written down. Screening rigour we figured out the hard way. Stories from engineers who joined frontier labs through the network — and the engineers who didn't and what they're doing now. Updated weekly.
// This week
What we published this week.
The flagship piece for the May 2026 issue plus the strongest companion.
// COMP & MARKETWhat the May 2026 frontier-hiring data actually says.
We dug into 5,800 senior-IC offers across 52 companies. Three numbers moved fast enough to change how you should plan the next 12 months — and one moved in the opposite direction from what every other comp tracker is reporting.
// SCREENINGThe one question we ask every senior IC panel.
It’s not the hardest one. It’s the one whose answer tells us the most about who you’d be on month four.
// Recent
The archive.
Filter by topic above, or search by keyword. New posts every Wednesday.
Comp · FlagshipWhat the May 2026 frontier-hiring data actually says.
5,800 senior-IC offers across 52 companies. Three numbers moved fast enough to change your 12-month plan; one moved the wrong way.
ScreeningThe one question we ask every senior IC panel.
It’s not the hardest one. It’s the one whose answer tells us the most about who you’d be on month four.
CompSigning bonuses became normal in 2026. Here’s why.
38% of senior offers we tracked in Q1 included a signing bonus above $200K. Last year it was 21%. The structural reason isn’t what you’d guess.
ResearchAgent-eval design is the new bottleneck role.
The labs are all building agents. None of them have enough people who can design the evals that catch when the agents fail in production. What that means for your next move.
ScreeningFour mistakes engineers make on the 13-day project.
We’ve reviewed about 2,400 of these. The mistakes that lose offers are remarkably consistent — and almost always about scope and write-up, not execution.
AgentsThe seven failure modes of production tool-using agents.
Schema drift, argument hallucination, looping, and four more. The taxonomy we use when we talk to candidates about agent reliability.
CareerWhen quitting your frontier lab is the right call.
Frontier-lab engineers reach a specific kind of plateau around year three. The signs are real and not the ones you’d guess. What the engineers who left and did better did differently.
CompIndia frontier comp climbed 34% YoY. Here’s what’s behind it.
Sarvam, Krutrim, and three quieter labs. Why the senior IC band in Bengaluru moved faster than any other geography this past year — and what it took to close the gap.
CareerThe resume rewrite that got 11 frontier-lab callbacks.
One senior engineer’s exact diff — before and after — and the three structural shifts every frontier-lab reader wants to see in the first 8 lines of a CV.
ResearchThe 2026 post-training canon. Twelve papers.
The reading list we hand to engineers ramping into a post-training role at a frontier lab. Twelve papers, in order, with what to actually take from each.
ScreeningHow we read a frontier lab’s hiring signal.
Headcount growth, equity refresh tiers, signing-bonus patterns, internal vs. quiet listings. The five signals we use to figure out a lab’s actual trajectory — public roles say very little.
CareerWhy we built a feature for moving with your team.
14 small groups have moved into frontier labs together since the network launched. The pattern that emerged surprised us — and reframed how we think about senior AI engineering hiring.
Wednesdays. One post. No noise.
Every Wednesday we publish one piece — comp, screening, research, or career — and that's the only email you get from us. 8,200+ AI engineers and hiring leaders subscribed.
// ONE EMAIL/WEEK · UNSUBSCRIBE LINK IN EVERY ONE