Vaibhav Gupta

Making a programming language - BAML. @boundaryml, 🦄 aitw podcast: YC, google, msft, deshaw, and other things

[Vaibhav Gupta @vaibcode]

I don’t enjoy the syntax and tooling we use to building AI pipelines. Langchain/ai sdk/et all have never really felt “right”.

No type safety, single language support, prompt is hidden until I buy some observability suite.

So we made a thing

https://t.co/eU06qqQHjK

[Vaibhav Gupta @vaibcode]

@velonxbt there's a reason professional athletes have coaches.

to go to the absolute limit you need someone observing you from the outside.

[Vaibhav Gupta @vaibcode]

Work as many hours as you can sustain. for most it caps around 80 hours / week. past it, is unsustainable.
if you're lucky enough to have it be 120+, do it.

we all have one life. don't let anyone shame you one way or the other.

[immad @immad]

You can work 5 days a week and succeed as a startup.

Mercury has done that from day 0 and we are valued @ $5.2bn 7 years after launch.

I have been an entrepreneur for 20 years and raised 3 kids while doing it.

The point of success is to have a great life not just a startup 😊

[Vaibhav Gupta @vaibcode]

I think that 80h is sustainable for most if they truly love doing what they do. at some point it doesn't feel like "work".

sadly, most folks don't find what they truly love to do or what they love to do isn't financially profitable. in that case, i recommend finding the highest financial reward for minimum effort, so you can do what you actually love.

[Vaibhav Gupta @vaibcode]

@Scrubghetti yea sadly macos hardware perf counters require sudo permissions, so i had to try, but luckily any actual data all lives in the cloud and only has readonly access.

it was a very successful experiment!

[Vaibhav Gupta @vaibcode]

i just gave claude sudo access cause i'm too lazy to run it manually... i'll report back.

[Vaibhav Gupta @vaibcode]

@RhysSullivan whats one low hanging thing that would make them better?

[Vaibhav Gupta @vaibcode]

@ossaijaD @benhylak @boundaryML all day every day

[Vaibhav Gupta @vaibcode]

anthropic's workflow feature bills even when you don't use it 😂

big fan of the new workflow feature, but now every other smaller task where i accidentally use "workflow" is just a smidge slower.

[Vaibhav Gupta @vaibcode]

@antoniosarosi codex make no mistakes!

[Vaibhav Gupta @vaibcode]

@mustafa_2vec Let’s go exponential

[Vaibhav Gupta @vaibcode]

I can't go back to not using worktrees... what was i doing my whole life.

[Vaibhav Gupta @vaibcode]

@yigitozgumus How big are we talking?

[Vaibhav Gupta @vaibcode]

@noname_oni @boundaryML oh this is artisinal deslopping! its task dependent and each engineer on the team has their own flow.

[Vaibhav Gupta @vaibcode]

Deslopify is the bottleneck.

[dex @dexhorthy]

lmao TIL @vaibcode has a "deslopify" column for tickets

[Vaibhav Gupta @vaibcode]

@MichaelThiessen Yea… I find some UIs are nice for this. Esp those that show merged icons 😅

[Vaibhav Gupta @vaibcode]

Hopefully it lives up! Baml is actually currently best for agentic tasks like taking in a csv and normalizing it or dynamic schemas based on what customers ask for. Your url example is a great one!

Best way is to use this agents.md https://t.co/G1zP8c2KCF

And just describe the problem to an agent and it’ll do the rest! Ask is to build some uis and add semantic streaming for extra wow.

[Antonio Sarosi @antoniosarosi]

At @boundaryML We're building building a full compiler, VM and async runtime.

Kinda like TypeScript + V8 + Node.

[Vaibhav Gupta @vaibcode]

@noname_oni @boundaryML what do you mean whats the prompt?

[Vaibhav Gupta @vaibcode]

@dexhorthy obviously its empty cause we don't write slop

[Vaibhav Gupta @vaibcode]

@ThePrimeagen and now we have 2 ways of doing things
map(array, func) vs https://t.co/JwmuauByfM(func)