Dev

Building world’s first A2A network @loop2ai | ex-AI infra @sprinklr | CS grad @iitroorkee

[Dev @debuggerdev]

“Meaning is not something you stumble across, like the answer to a riddle or the prize in a treasure hunt. Meaning is something you build into your life.”

[Dev @debuggerdev]

@ThePrimeagen 200 commits/day, feels like we’re optimizing for code generation

[Dev @debuggerdev]

@GergelyOrosz everything is ‘AI-powered’

[Dev @debuggerdev]

@pmddomingos so AI becomes the generalist

[Dev @debuggerdev]

You’re not "choosing to move" through spacetime
You are a path through spacetime

[Andrej Karpathy @karpathy]

This is the the quote I've been citing a lot recently.

[Dev @debuggerdev]

@karpathy @VBarsoum cozy codin, aka being together but the third person is the model

[Dev @debuggerdev]

treat agents like a team, not a swarm

you don’t want 5 engineers editing the same file at once
you want:
one person writing
others reviewing, poking holes, suggesting directions

multi-agent works the same way
parallel insight
serialized decisions

[Dev @debuggerdev]

@mattpocockuk it will need having enough design intuition to pick the right ones

[Dev @debuggerdev]

@mattpocockuk you define the solution around services, LLM builds it perfectly but the real problem needed state as the primary concept

[Dev @debuggerdev]

@ThePrimeagen if improvements require degrading the past, that’s not progress

[Dev @debuggerdev]

@claudeai rigor is improving

[Dev @debuggerdev]

@GergelyOrosz AGI for everyone until the invoice

[Dev @debuggerdev]

@GergelyOrosz classic AI upgrade:
more power, less usability

[Dev @debuggerdev]

@Polymarket AI reduced typing, but increased overthinking

[Dev @debuggerdev]

exploring local agent setups (openclaw + local models)

feels promising
but unsure how it holds up beyond demos

anyone here running it for real workflows?

[Dev @debuggerdev]

@DavidKPiano AI doesn’t fail randomly, it fails outside familiar patterns

[Dev @debuggerdev]

@PeterDiamandis the interesting part is what gets amplified

cheaper AI doesn’t just increase experiments
it increases bad experiments too

so iteration speeds up
but so does noise

the real advantage shifts to
who can filter faster

[Dev @debuggerdev]

LMs are great at knowledge, but they struggle one layer above it

judgment
what will actually work
what’s just plausible but wrong

i’m noticing this a lot

they can generate many approaches
but don’t really know which one is worth pursuing

that part still falls on us

[Dev @debuggerdev]

spent less time coding today

but way more time deciding

which AI output to keep
which to throw away
what actually makes sense

feels like the work didn’t go away

it just moved from typing
to judgment