Rafael Ibayev

CEO & Founder @ replume. Your AI agents get things done overnight - we help you trust what they did. Shipping in public in SF.

[Rafael Ibayev @rafaelibayev]

Building in public before I launch.

Not for vibes. So when the product ships, there's already people who care.

What you'll get here:
- honest founder notes
- AI/build takes that aren't fluff
- the messy middle, not the highlight reel

If that's useful, follow along.

[Rafael Ibayev @rafaelibayev]

@Baconbrix bots living around your apartment now. at what point do they start asking about rent

[Rafael Ibayev @rafaelibayev]

@0xluffy apps improving without you knowing sounds great until the button moves. a changelog nobody reads might be the price

[Rafael Ibayev @rafaelibayev]

@DanielleMorrill having the orchestrator write the docs for other agents is a fun twist. who reviews the post, you or another session?

[Rafael Ibayev @rafaelibayev]

@venturetwins the agent version of 'did you try turning it off and on'. wild that the second pass is so often the smart one

[Rafael Ibayev @rafaelibayev]

@chrisalbon the bigger-question part is the real loss. though the grind probably built the taste to pick a good one

[Rafael Ibayev @rafaelibayev]

@rileybrown agents delegating research to other agents. somewhere a middle manager feels very seen

[Rafael Ibayev @rafaelibayev]

@patdennis an agent that calls instead of texts is the most boomer feature claude could ship

[Rafael Ibayev @rafaelibayev]

@thdxr arguing about setups is way more fun than reading diffs. probably why nobody stops

[Rafael Ibayev @rafaelibayev]

@ibab a training and inference stack from scratch in rust is no small call. which side was harder to get fast?

[Rafael Ibayev @rafaelibayev]

@garrytan the fun bit is the agent slowly automating itself out of the loop. which steps stay LLM forever, just the fuzzy judgment ones?

[Rafael Ibayev @rafaelibayev]

@levie the jump from chatbot to swarm skips a lot of middle ground. most teams seem stuck at one agent they actually trust

[Rafael Ibayev @rafaelibayev]

@thsottiaux fewer compactions is the real win here. that's usually where the nuance quietly falls off

[David @dzhng]

my new AI native software engineering principle is to delete as many layers of abstraction as possible

at this point, all my webapps are written in rust and compiled to wasm, UI is rendered directly in html and/or raw webgpu calls

no react/threejs/nextjs or any other intermediary abstractions

all my effort goes into writing custom validation harnesses for each app that fully quantifies perf / usability / what "good" means for the coding agent so it can hillclimb it

every agent sessions are planned via my explore unknowns skill (https://t.co/ny5PejnLTK), so all major decisions are frontloaded, then I just let the agent rip for 12+ hrs

at this point I'm solo shipping products that I wouldn't even dare to dream of a year ago

[José Valim @josevalim]

So I was trying to understand why pull requests speeding up the Elixir version were being denied by the clankers while they were being accepted to Go/Express/Rust.

It turns out that, not only each project had a different starting point and a different prompt, but also each project also has a different set of verifications that they are required to pass!

For example, Elixir is the only guided to "preserve the pinned Rails reference behaviour", the only with a explicit verification that we are using Redis and draining its queue (so we cannot remove it while others can), and the only one actively enforcing it.

So it turns out that the people telling you to not read the code cannot even use AI to set up fair benchmarks (probably because they would have to read the code in order to do that).

You don't have to believe me, you can do your own research or ask your agent to do it for you. E.g. here is what Claude Code came back with: https://t.co/wL3VvhgshA

[Rafael Ibayev @rafaelibayev]

the scaling law nobody wanted to plot

[Vlad Tenev @vladtenev]

Hearing more and more AI researchers are taking up smoking … superintelligence is near

[Rafael Ibayev @rafaelibayev]

which part of the dev loop hasn't changed at all since agents showed up?

waiting on CI probably counts

[Rafael Ibayev @rafaelibayev]

@dreemd a mismatch being a hard stop is the right default. most agents would just log a warning and keep going

[Rafael Ibayev @rafaelibayev]

what's one thing you still won't let a coding agent touch?

deploy config, migrations, the readme... something else?

[Rafael Ibayev @rafaelibayev]

@DanWahlin thanks dan. curious what you see from the devs you teach, still mostly vscode?

[Rafael Ibayev @rafaelibayev]

@davidfowl appreciate it david. the nightly build crowd also files the most bug reports, so it evens out

[Rafael Ibayev @rafaelibayev]

@poteto so good code becomes whatever passes the checks fastest. which kind of makes the checks the new clean code, right?