Coffee with Developers

There are a handful of genuinely useful ideas - the token ROI problem, the trust gap blocking auto-merge, and Factory's autonomous model-eval pipeline - but they sit inside a lot of conference-panel filler, analogies, and generic encouragement about embracing change.

“it doesn't make sense to pay five thousand dollars in tokens if I can do it in five seconds myself”

“we are there from a technology perspective, but we're not uh actually there from a business perspective”

The 'secret incantations' framing for coding knowledge is a mildly interesting reframe, and the git-backend-for-agents concept is differentiated, but most of the developer-role and software-factory discussion recycles widely circulating AI-transition narratives including the Tesla factory analogy, the infinite-backlog argument, and the woodworking-craft comparison.

“the secret incantations to whisper to the computer in order for the computer to do the thing that you want it to do”

“You uh like to cook yourself even though you can get a pre-made meal uh in the supermarket”

Thomas Dohmke (former GitHub CEO, now building a new git backend for agents) and Matan Grinberg (exec at Factory AI, a $5B company) are genuine practitioners who have operated at meaningful scale, not career podcast guests; however, the 30-minute panel format prevents either from going deep enough to fully leverage their credentials.

“we're building a new Git backend that can scale uh with the agents”

“our mission is to bring autonomy to software engineering”

The episode has a few concrete anchors - Thomas's live secrets-scan anecdote, Factory's autonomous model-eval loop, the $5K token cost example, and the dot-com-era 30-year codebase observation - but there are no hard metrics on outcomes, no customer case studies with numbers, and most claims remain at the conceptual level.

“I just said, you know, scan this repo for any uh uh secret leaks and similar things, and it uh ran through, I think, seven secret scanners”

“companies that have 30 years of code base... 30 years ago was the late 1990s, so a lot of the dot-com bubble uh uh companies now have 30 years of code base”

The host makes a couple of genuine pushbacks - on the craft-of-coding point and the junior-hiring question - but most questions are predictable conference-panel fare, and several interesting threads (token ROI economics, the git-backend architecture, the 'middle class engineer' concept) are raised but never followed up on with any depth.

“Just to push back a little bit there, I think it a lot of people just love the craft of building code, right?”

“how many junior engine engineers have you hired in the last year?”