Alex Sofroniev

🚢 - Automated Docs Generation

[Alex Sofroniev @alexsofroniev]

I'm 24.

Solo founder from Bulgaria living in Valencia.

Looking to connect with more builders & indie hackers!

[Hauke 🌱🦀 @haukejung]

I'm 38.

Solo founder from Germany.

Looking to connect with more builders & indie hackers!

[Alex Sofroniev @alexsofroniev]

@StevBuilds @X builders and small teams shipping fast.

AI helps the person writing the code.

it doesn't help the user who opens your docs and finds last month's screenshots.

code moves faster now.

so docs fall behind faster too.

[Alex Sofroniev @alexsofroniev]

@kocienda That multiplexer is the part that makes me want to try it.

Starting with one great integration beats five half-working ones.

If Tug ever opens that backend up to other agents, I'd love to wire mine in.

[Alex Sofroniev @alexsofroniev]

@rufo_ai Docs that fall behind the product.

You ship a feature, the README is already wrong, and new users bounce because they can't get started.

That's the problem ShipDocs is built for.

[Alex Sofroniev @alexsofroniev]

@metahacker_ Ten years of copy is an unfair advantage.

Most people hand that part to the model and hope.

You already know what good looks like, so the agent is just faster fingers.

[Alex Sofroniev @alexsofroniev]

@StevBuilds @X ShipDocs.

Docs for products that ship faster than anyone can write them.

I got tired of features going out undocumented, so I built the thing I kept needing.

[Alex Sofroniev @alexsofroniev]

@AkhaniVishant fair, that's the boring answer and probably the true one.

doing both every day it is.

[Alex Sofroniev @alexsofroniev]

@rufo_ai right here, honestly.

replies like this one.

no ads, no launch yet. just showing up in threads where people talk about the problem.

[Alex Sofroniev @alexsofroniev]

@suraj_sharma14 yes, the threshold is the part that matters.

"usage dropped" doesn't tell you much on its own.

"usage dropped 40% and renewal is in 30 days" is something you can actually act on.

[Alex Sofroniev @alexsofroniev]

@rodri_shema exactly.

the meter changes how you work.

when every run costs you, you babysit the agent.

when it doesn't, you just let it run and check the result.

[Alex Sofroniev @alexsofroniev]

@Astrodevil_ underrated exercise.

the agents mostly pick whatever they can understand fastest.

that makes your README and docs part of distribution now,

not an afterthought.

[Alex Sofroniev @alexsofroniev]

@notathrv none of the three, honestly.

Pi + DeepSeek on a flat-rate plan.

for a solo dev, predictable cost beats the best benchmark.

[Alex Sofroniev @alexsofroniev]

@evvvvince weeks like this are why I stopped tying my setup to one vendor.

the harness stays.

the model behind it can change.

boring setup, fewer surprises.

[Alex Sofroniev @alexsofroniev]

@sermakarevich saving this.

the harness matters more than people think.

same DeepSeek model feels like a different tool depending on what wraps it.

permissions is the part I'd read first.

[Alex Sofroniev @alexsofroniev]

@read__the__diff this is painfully real.

I'm on multiple flat-rate coding plans at once.

picking the model is easy.

remembering which key lives in which harness is the actual job.

[Alex Sofroniev @alexsofroniev]

@tusharmath counterpoint: I live in the terminal on purpose.

Pi + DeepSeek, no IDE.

the babysitting isn't the terminal's fault.

it's whether you trust the diff.

[Alex Sofroniev @alexsofroniev]

@ayowandde exactly.

the click is cheap.

cleaning up after an agent that guessed wrong isn't.

[Alex Sofroniev @alexsofroniev]

@Isaac_Dev0 Yeah, under $100 MRR every subscription has to earn its spot.

The coding plan is the only one that pays for itself the same day.