← sources

Perspective from inside a very AI Pilled start up

source: reddit (reddit:cscareerquestions)published 2026-10-21 12:16ingested 2026-10-22 02:37

retained text

My goal here is to give perspective into how people are working when there is an incredible amount of work to do and very little restrictions on AI use. Where I work I would say we are in the top 5% or so of AI adoption but not in the top 1%.

Context:

* "Hard tech" software - i.e. software much more complicated than normal SaaS but not a 10/10 hardcore. Think similar industry to EDA software

* Not an AI company but there seems to be rabid demand from enterprise for our AI product... It seems to me that a lot of enterprises have crappy access to AI so there is a lot of value in bundling an agent since they can't bring their own agent as easy as us

* Roughly 8 engineers who are contributing code full time, while other people drop somewhat slopped PRs as needed

* All the engineers have several $200 subscriptions to their preferred provider and can get more subscriptions as needed

* 25-30 PRs to the mono repo per day, currently the codebase grows at ~25% per month in terms of net new lines. Also lots of the existing codebase gets refactored every month so something like ~35% total code turnover 

* Code review is the bottlekneck. People try and not just auto approve everything and we use lots of AI code review tools as well. Obviously people are not reading all the individual lines of code but I'd say we are still spending lots of time on architecture. Sometimes too much time 

* In my first ~2 months I added around 500k lines of code, much of that was tests/ tooling/ CI so not all of this is like production code

* I don't use an IDE anymore and neither do most of my colleagues. I do use a diff tool with LSPs enabled to review my code and peoples PRs... so pretty close to IDE just not for development per se

* Overall it is going well - we are iterating fast and close to closing some potentially huge enterprise deals. Outside of these big fish lots of smaller contracts that are over the line

* There is some incidents but we are getting them down to a reasonable amount. Recently there was a very serious incident that put us on thin ice with one of our largest customers 

* We have had the most luck on the "verifiable" portions of the codebase i.e. algorithms and workflows that we can clearly benchmark. Several cases of finding exponential speedups in our core workflows that wouldn't have been possible

* Our "algorithmic core" is getting really strong because AI is letting the people who were already good at math on our team do incredible things and letting people like me who know very little fix roadblocks 

* Our UI/ UX is really, really bad - there is ten different ways to do things, huge amounts of feature creep and the happy paths are painful. I blame this partially on AI/ vibecoding and partially on skill issue on our teams part 

* In terms of contribution to the business there is clearly "skill" to using AI atleast currently. The gap between most productive and least productive engineers is very large in terms of end contribution (not just PRs or LoC etc)

* At this point literally 0% of the codebase is human written

* We are creating increasingly restrictive lint rules and gen code patterns so that agents stop reinventing the wheel in every PR... this has been mostly successful

* Team is fairly young but org is interested in mostly hiring seniors going forward 

* If someone told us they didn't use AI heavily we wouldn't consider hiring - the default assumption is that you are a heavy AI user here

* Work generally expands to fill the time and people work a ton, most of the employees have SO's/ exercise routine/ social lives so not full burn out schedule

So anyway it is a mixed but clearly positive bag. We are moving atleast 3x the speed of what would've been possible before but realistically more like 5-7x I would guess. It is mostly working in terms of helping us get to product market fit and then shipping super fast. 

I often disagree with the attitude I see on this sub regarding actual productivity gains. We are not really a "greenfield" since we have plenty of active customers but not legacy code either. Hope this is helpful report from the front.

document fetched 2026-10-22 00:24 · canonical https://reddit.com/r/cscareerquestions/comments/1xbljiy/perspective_from_inside_a_very_ai_pilled_start_up