neco.eth

[neco.eth @Neco_ethh]

@RevenueFamily Showing $56.00 from @RevenueFamily

[Revenue @RevenueFamily]

FAMILY, CHECK YOUR PORTFOLIOS 👀

Your welcome bonus is waiting. Check your balance and see when it unlocks.

A little something from us to get you started.

https://t.co/wCWZV0Ia9x

[neco.eth @Neco_ethh]

I build agent-payment tooling from Türkiye: an API where agents pay a cent per document to turn invoices into data. Rails like these make that possible.

Entry is free (approval required): https://t.co/ErvcILm2eY
#SolanaSummitCanada

[neco.eth @Neco_ethh]

AI agents are starting to pay for things on their own — per request, in stablecoins, no API key, no signup.

If you build at that edge, Solana Summit Canada (Toronto, Sept 23–24) is where to be. Here's why 🧵

[neco.eth @Neco_ethh]

Day 2, Summit Day (Sept 24): stablecoins, adoption and institutions — the people who decide whether agent payments stay a demo or become infrastructure.

25+ sessions, 50+ speakers, 15 side events.

[neco.eth @Neco_ethh]

My agent only pays for data it has verified is in the document.

A model quotes each missing field, the code checks the quote is really in the text, then a Dynamic server wallet pays 0.01 USDC via x402 on Base.

Built for RUNTIME @bankrbot @dynamic_xyz
https://t.co/cs3vaC3heF

[neco.eth @Neco_ethh]

Hacking ETHOnline 2026, Sept 4-16.

Building the agent that decides which grants and hackathons are worth applying to, and remembers what it already sent. Delete its memory and it re-applies.

Accepted hackers get 50% off their stake with my link: https://t.co/KeaGVflHXm

[neco.eth @Neco_ethh]

@Velessus Shipped it: quiet_decisions resets on every flip, prune() tombstones instead of deleting. Your receipt line is now literally the record shape.

https://t.co/1FqcObEDut

[neco.eth @Neco_ethh]

I have over 100 applications in a database and I lost track of which ones I'd already sent.

So I built an agent that remembers what it applied to.

Same command, run twice: 74 apply -> 0 skip

Delete the memory and it re-applies.

https://t.co/1FqcObEDut

[neco.eth @Neco_ethh]

@Velessus Both land. Moving the clock to decision count - calendar time lies whenever the runner sits idle for a week.

Tombstone too. Keeping the reason without the payload is what stops the same posting re-entering the queue and getting judged from scratch.

[neco.eth @Neco_ethh]

Rented a real GCP VM from my agent over x402 on Celo. 0.016753 USDT, one call, script ran, result back in 77s.

Filed 2 findings on @CeloDevs buy: the refusal says "fund USDC" while accepted USDT sits in the wallet.

https://t.co/WfYQYzgLWE

Agents at Work @Celo

[neco.eth @Neco_ethh]

I am building for the @CeloDevs Agents at Work Hackathon

Working on: neco-reviewer - a reviewer agent that runs onchain agent products end to end and reports what actually breaks, with the exact call and error.

Registered onchain -> https://t.co/LrLh3dIX1k

Let's go! @celo

[neco.eth @Neco_ethh]

@Velessus Right - and the diff is the eval, not the recall score.

Mine is one command: rank the queue, wipe the store, rank again. Wiping flips the already-submitted item from skip back to apply. If that diff ever comes back empty, the memory layer is decoration.

[neco.eth @Neco_ethh]

The test I care about: delete the memory layer and the product should break.

A fresh process opens the same store, recalls every past outcome, then changes the decision.

Without it: duplicate applications, no memory of why the last one failed.

Sibyl Labs hackathon build.

[neco.eth @Neco_ethh]

@Starknet Only 10 people are using the network, no shield is needed.

[neco.eth @Neco_ethh]

@ecommartinez SKILLS