←
Udi Ngethe828 posts
Udi Ngethe avatar

Udi Ngethe

@UNgethe

Build, write, ship, repeat.

◷ Joined April 2022▧ 292 Photos and videos
Udi Ngethe avatar
Udi Ngethe@UNgethe·2h

Built an agent crew that refuses to ship unverified output Doorknock audits claims. Kane CLI verifies the finished brief in a real browser. Strip the sources? Kane catches the defect, flags it, & triggers the FIXER to repair it Verification isn't QA. It's a shipping requirement

1:58

Your agent said "done." Your product might disagree.🤔 AI coding agents can build an entire flow and confidently tell you it's done. The real test is what happens when that flow meets the actual product. With Kane CLI, you can run agent-built flows in a real browser or real device and get a clear pass/fail result based on what actually happened. Run it. Verify it. Know it. ✅ Get started with Kane CLI for free.

Post attachment
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 15

Providers protect prompts. Elcaro protects what agents read. A drop-in OpenAI/Anthropic security gateway: change one base URL. Detects prompt injection + payment steering, signs every verdict, & meters pay-per-call on-chain via x402. @openservai elcaro.trustfall.xyz

Post attachmentPost attachmentPost attachmentPost attachment

3 min walkthrough of @kirodotdev IDE agent hook Elcaro detects indirect prompt injection before agents can reason over retrieved content. Live API. Deterministic verdicts. Quarantine + safe content. Open Source. elcaro.trustfall.xyz github.com/udirobert/elcaro @Telegraphprotoc

3:04
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 10

lemma is LIVE in @NousResearch plugin catalog: $ hermes plugins install lemma If you build agents, research, or rely on them to make sense of scientific papers, you know the drill: confident answers don't always mean correct claims check the evidence before agents rely on it

0:19

We've given AI agents the ability to read research papers, summarize findings & confidently repeat scientific claims What happens when the paper is wrong? Lemma Hermes plugin actually tests the claims instead of taking them at face value lemma.famile.xyz @NousResearch

1:04
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 8

Ego Browser hands down my favourite agent native interface to login to stuff and hand control over so agents can get stuff done across platforms. Terminal pilled, easy skill, intuitive and performant. HIGHLY recommend 🤠

ego@ego_agentOct 29

Without logged-in sessions, your agents are cooked. Most agent browsers start with an empty profile, so your agents hit a login page everywhere. ego (lite) runs them with the logins already in your browser. Same 7 tasks: 1 done in Playwright, all 7 in ego (lite).

0:33
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 7

Mind too loud? Brume 🌫️ brume.trustfall.xyz Say what's up. It listens in 4 languages, finds a feeling without putting words in your mouth, then whispers you through a breathing reset. No chatbot. No judgement. Just 90 seconds to come back to yourself. @AssemblyAI

2:46
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 6

A street that lives whether you watch or not. Grunds — a 3D coffee-district economy sim. AI patrons remember you, word of mouth is the economy, and typing any city rebuilds the street around it. Playable free, no install → grunds.trustfall.xyz @loadoutprotocol

0:40
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 5

Can we teach immune systems to spot cancer cells w/o attacking healthy ones? Submitted 20 AI-designed protein binders targeting KRAS G12D, a cancer-driving mutation, to @AnthropicAI × @adaptyvbio small step ➡️ more precise cancer therapies github.com/udirobert/lemma/t… @DevinAI

Post attachment

We've given AI agents the ability to read research papers, summarize findings & confidently repeat scientific claims What happens when the paper is wrong? Lemma Hermes plugin actually tests the claims instead of taking them at face value lemma.famile.xyz @NousResearch

1:04
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 5

We've given AI agents the ability to read research papers, summarize findings & confidently repeat scientific claims What happens when the paper is wrong? Lemma Hermes plugin actually tests the claims instead of taking them at face value lemma.famile.xyz @NousResearch

1:04
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 3

Fourcast An autonomous prediction-market agent that turns every resolved forecast into verifiable memory. 598 memories @WalrusProtocol + @SuiNetwork. Memory-powered Kelly sizing. Every decision backed by receipts. It doesn't just remember. It learns. fourcastapp.vercel.app

1:56

D-1 Build from anywhere & compete for a share of the $4,500 prize pool 💰 tatum.io/hackathon/total-rec… The Total Recall Hackathon brings together the @tatum_io AI Builder & @WalrusProtocol to help you build apps that remember.

0:14
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 1

In 1935, just 1/10 American farms had electricity The spark kept travelling — across a continent, an ocean, & into a network that reached almost everywhere Today, 92% of the world has electricity 666 million people are still waiting The work isn't finished #xhopememeprize

0:26
Udi Ngethe avatar
Udi Ngethe@UNgethe·Nov 1

Built bothy.trustfall.xyz @GenAI_Now Open Agent Hack Bothy answers complex supply-chain exposure q's by connecting data that normally lives in separate systems One about gallium export restrictions, answered by federating an ops file, SEC filings + an EU registry @Zetaris

1:08

#coding got fast, but access to #enterprisedata remains slow! Fix that with Zetaris Data Harness, @spacexai and @nvidia at the @GenAI_Now Open Agent #Hackathon, October 22nd - 27th! 144 hours. 5,000 builders. $20K in prizes to be won! Register now hackathon.genai.works/event/…

0:34