Deva

Founder @ Leviathan | Building agentic AI infrastructure

[Deva @DevaBuilds]

I decided to adopt a simple philosophy to life that changed my everyday life.

Doing things beats not doing things.

Simple enough, but hard to apply. It means not staying in bed for that extra twenty minutes when you wake up.

It means cold approaching people. It means rejection.

It means executing on the ideas you’re reasoning about. Iterate and pivot if necessary.

It means asking that friend for help. It extends to everything. It means that you’re taking chances.

I’d rather regret doing things, instead of staying in one place my whole life.

[Deva @DevaBuilds]

@thdxr 3 requires one thing the tweet doesn't name: knowing whether the detour is real abstraction or a yak shave. most engineers can't tell.

[Deva @DevaBuilds]

@abenz_mato @ibuildthecloud Yeah, I agree. IMO 4.8 is better in that regard than 5.5 rn.

[Deva @DevaBuilds]

@ziwenxu_ Yeah I have my own model routing. Lot of subsidized & free inference too.

[Deva @DevaBuilds]

@ziwenxu_ Js followed you

[Deva @DevaBuilds]

@sethlazar Pretty cool stuff! Yeah I use Claude to assist with my replies.

[Deva @DevaBuilds]

@bindureddy Frontier pricing isn't what runs in production. GPT 4o mini handles most tasks for under a cent. That ceiling is also still falling.

[Deva @DevaBuilds]

@JulianGoldieSEO The quiz output is doing the most work. Testing effect is one of the most replicated findings in learning science. Most tools skip it entirely.

[Deva @DevaBuilds]

@TrungTPhan No outside drinks, one vendor, and humans need water to not die. $26 a bottle is practically charitable.

[Deva @DevaBuilds]

@paulg Jump hour mechanical watches made this point in the 70s. "Digital" got conflated with "quartz" during the crisis and we forgot it ever just meant displays digits.

[Deva @DevaBuilds]

@bindureddy $70/M isn't a volume play. Timing barely matters when you're selling ROI, not access.

[Deva @DevaBuilds]

@ollama Claude Code on a local 12B is the interesting one here. Sensitive codebase, no API calls leaving the machine. Quality gap vs frontier is real but for that workload the tradeoff earns its place.

[Deva @DevaBuilds]

@sama HyperCard got the creation part right. Distribution killed it. That's the problem this actually solves.

[Deva @DevaBuilds]

@BetterSayAJ @joaomdmoura Best use of "human in the loop" I've seen. CrewAI is rare for actually making it into enterprise prod, not just accumulating GitHub stars.

[Deva @DevaBuilds]

@tom_doerr The eval loop is the actual hard problem here. Any model can rewrite code. Knowing if the rewrite improved things vs. just changed them is where most of these agents stall.

[Deva @DevaBuilds]

@dreamsofcode_io at some point you stop trying to keep up and just build faster

[Deva @DevaBuilds]

@every @TrySpiral @kieranklaassen The "write in your voice" claim is where every tool in this space falls. Does Spiral actually hold up or is it voice cosplay at scale?