The AI Record
We are definitely still bottlenecked on reviews, especially for things that are touching some architecture pieces. And it's actually more subtle than just being bottlenecked on review, … we can carve out time differently. It's … bottlenecked on human ability to even fully conceptualize what we're doing.

Mike KriegerHead of Anthropic Labs · Leader, Anthropic

Context. Answering a question from the host on whether code review is a bottleneck and whether code could be merged without a person looking at it.

Interview: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 10:13.

Checked against the source [ Grade C ] Said in the last year Anthropic: $47 billion revenue, run-rate, May 2026 Sells AI tools

Filed under

Who is speaking

Mike Krieger, Head of Anthropic Labs. Works inside Anthropic Labs and was Anthropic's chief product officer, so he describes how the team builds and reviews work from the inside. Anthropic sells AI products, so his account of its own practice is first-party.

Anthropic sells AI or agent tooling, so it has a reason to publish flattering numbers about AI. The passage stays in the archive; read it with that in mind.

How this was checked

Checked against the source by a first reader, then confirmed independently by a second reader who fetched the source without seeing the first reader's notes (Independent review agent (Claude Sonnet), blind second pass from passage text and citation only, no repository access, 2026-10-31).

Matched in the whisper transcript, Krieger answering the code review and Git question. The … marks omit 'because that's OK.' and the words 'on human' are in the source as written. Wording is otherwise exact. Timestamp not checked against 10:13.

How to cite this passage

Mike Krieger (Anthropic), “How Anthropic Builds: Lessons from Labs”, AI Engineer (YouTube), 2026-07-02, at 10:13. https://www.youtube.com/watch?v=qqrk7CtkuIw Archived at https://the-ai-record.com/passages/anthropic-labs-bottlenecked-on-review-human-understanding-2026

More from the same source

if you align the org chart too much to the individual projects, you're going to end up re-orging every two weeks, which would be a total nightmare. … And there is the concept of a bet lead or a directly responsible individual. But the interesting thing is that they don't manage usually any of the other people, which … breaks the … previous way in which a lot of these things were done. But I think it leads us to be really flexible when you say, OK, actually, this project is not going to work out. Let's disband and keep going, and it's not a big deal.

Mike KriegerHead of Anthropic Labs · Leader, Anthropic

Context. Continuing his answer on how Anthropic Labs is structured; 'bets' are the project teams inside Labs.

Interview: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 12:46.

Checked against the source [ Grade C ] Said in the last year Anthropic: $47 billion revenue, run-rate, May 2026 Sells AI tools Passage page

I was like, OK, I need to port this whole thing from Python to TypeScript. Like, as a-- you know, if I put on my, like, 2010s engineering hat, or even my early 2020s, like, that's a dumb idea. … But I was like, I think this is doable now. And I basically created this dynamic workflow set up. And over the weekend, … port the whole thing, like, verify it, double check it, … and then came back Monday to complete a workflow that was a ported version of that thing.

Mike KriegerHead of Anthropic Labs · Leader, Anthropic

Context. Describing a Labs project he moved from Python to TypeScript, in answer to a question on being more ambitious with prompts.

Interview: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 4:38.

Checked against the source [ Grade C ] Said in the last year Anthropic: $47 billion revenue, run-rate, May 2026 Sells AI tools Passage page

But especially in a labs-type group where our whole cadence is two-week reviews where every project goes up for, we call it persevere or pivot. So basically, every project is up for review, and either it's time to keep going, … or it's time to pivot it, or even shut down. And we shut down projects basically every single one of those cycles. And the more you do it, the less it's just like, oh, no, my project is shut down. I failed. It's like, no, that is definitely the intention of the lab team is to prototype quickly, try to ship internally, maybe get it to early access, and if it doesn't work, wind it down.

Mike KriegerHead of Anthropic Labs · Leader, Anthropic

Context. Answering how Anthropic Labs is structured.

Interview: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 12:11.

Checked against the source [ Grade C ] Said in the last year Anthropic: $47 billion revenue, run-rate, May 2026 Sells AI tools Passage page

actually talked to Claude about the code and say, all right, these are the questions that I would have. Can you go investigate it? So it is kind of … code review, but still human-driven, and for the really important ones. And for the ones that are cosmetic visual changes, it's much more like we'll fix forward if we need to fix forward.

Mike KriegerHead of Anthropic Labs · Leader, Anthropic

Context. Continuing his answer on review. The sentence before this passage, on not reviewing every line of a pull request, is not quoted because the two passes disagreed on its wording.

Interview: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 11:13.

Checked against the source [ Grade C ] Said in the last year Anthropic: $47 billion revenue, run-rate, May 2026 Sells AI tools Passage page

We have one of our Slack challenges, like Project Unship, which is like, what is in the product right now? … Like, we unshipped styles, I think, recently, … used by a small percentage of people … very, sort of, prescriptive in the way that it worked, and skills were a much better application, something like that.

Mike KriegerHead of Anthropic Labs · Leader, Anthropic

Context. Answering a question on what Anthropic would delete from its products.

Interview: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 15:43.

Checked against the source [ Grade C ] Said in the last year Anthropic: $47 billion revenue, run-rate, May 2026 Sells AI tools Passage page

But most usage is actually much more delegating via tagging… And then more advanced version is really trying to think of it as a teammate that is actually … has memory and can be proactive. And that's just really changed how we operate internally. It's much more like this multiplayer async proactive way than it is most people off in their own CLIs.

Mike KriegerHead of Anthropic Labs · Leader, Anthropic

Context. Describing how people at Anthropic work with Claude. The recording's transcripts render the name of what is being tagged unclearly, so the passage is cut at 'tagging'.

Interview: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 8:57.

Checked against the source [ Grade C ] Said in the last year Anthropic: $47 billion revenue, run-rate, May 2026 Sells AI tools Passage page

So one of the reasons we built … artifacts that we shipped a couple of weeks ago was partially for that, which is you would send somebody a PR, and then they'd be like, I don't know, man, this is like 2000 lines of code. It looks like code to me. And what we started doing instead is sharing much more like, here's a … artifact. Here's the explanation. Here's the intention of the change. Here's the trade-offs that were made.

Mike KriegerHead of Anthropic Labs · Leader, Anthropic

Context. Continuing his answer on review bottlenecks. The product name before 'artifacts' is rendered unclearly in the recording's transcripts, so the passage is cut around it.

Interview: How Anthropic Builds: Lessons from Labs, AI Engineer (YouTube), July 2, 2026, at 10:30.

Checked against the source [ Grade C ] Said in the last year Anthropic: $47 billion revenue, run-rate, May 2026 Sells AI tools Passage page

Something wrong? Report a correction. JSON: /api/v1/passages/anthropic-labs-bottlenecked-on-review-human-understanding-2026.json