I wanted an AI agent that could trade like me, so I built it. prototype is currently in testing. just threw together a basic landing page for it, will link below.

Introducing @ggbots_ai . It's a platform to build & deploy AI trading agents. Easily customize:.- market data your agent reviews.- trading strategy it follows with emotionless discipline.- risk guardrails & exchange connections. cc @_openeconomy

roast my pitch deck. I’ve got 24 hours to make it better. this is the current state of the ggbots-dot-ai deck for Vibe Code Camp. I need fast, brutal feedback. don’t hold back. what’s confusing? what’s cringe? what’s missing?. I’ll be iterating all day. 🧵👇.

one day left. no sleep, can’t stop, gotta finish. here’s what’s left for ggbots to be ready:. 1. finalize UI.2. craft the pitch.3. record the demo. follow along to see if I pull it off.

Vibe coding from the hawker center . Gotta hustle to submit something great for Vibe Code Camp in 3 days! . @_openeconomy

vibe code camp got me to *finally* stop grinding in a vaccuum and start actually posting about what I'm working on.

so far, this new approach has completely solved the clustering problem. no more defaulting to 68 every time. I'll share an update in a couple weeks, hopefully showing it actually improves accuracy this time! 🤞.

LLMs also have inherent bias, preference for certain numbers, evidently human bias present in the training data. running an improved version that forces the LLM to explicitly weight the relevance of the various indicators and calculate a final score from a pre-defined baseline:

AI gave nearly every signal a confidence score of 68%, it was so bad. the problem turned out to be "offset bias". basically there were so many data points that the AI could justify nearly any signal, ignoring clear red flags:

i pay over $1k/month for a premium trading indicator. it's ~90% accurate but i thought AI could make it even better. after building a sophisticated filtering strategy using reasoning LLMs cross referencing 14 other indicators. accuracy dropped to 80% 💀.🧵

wow earlier today I was like "I wonder if zkProofs could unlock a copy-trade proof decentralized trading bot?" then this drops. .

RT @gregisenberg: i miss the old facebook. not because i'm nostalgic for 2008, but because it solved a problem we're desperately trying to….

RT @balajis: An important kind of social network will be one where no bots whatsoever are allowed.

RT @MontyMerlin: I’m incredibly proud and excited to share that the Prosperity Pass Beta is live today! 👥🌐👥. A new chapter for onchain iden….

RT @regen_network: ✅ Proposal 50 Passed (100% YES). Approves the Regen v6.0 software upgrade, which includes:. • Cosmos SDK v0.47.• CosmWas….