Gabe

GTM Engineer · Forward Deployed AI Engineer

Hi, I'm Gabe. For 15 years, I've worked in marketing at the point where how a product works meets why it matters. Over the last five years, I've brought that perspective to GTM engineering and applied AI.

That means I care about clear sentences, good product taste, and systems people can use. Lately, I've been doing agent plumbing: finding what slows a team down and building the AI systems that fix it.

My working set includes Python, SQL, RAG, embeddings, vector databases, MCP, agentic systems, evals, context engineering, TypeScript, JavaScript, and the APIs around them.

How I think about work and AI: this video, this tweet, this doc, and this one on GTM.

Project	Stack	What it does
Tixmancer		AI agent that finds overlooked secondhand items and checks asking prices against your limit, working toward buying below market value.
Hedwig		Shared, revocable roles for related Solana apps: each app authenticates its users or agents and checks their access onchain.

Project	Stack	What it does
prospector		Forkable outbound lead pipeline: two scrapers, Clay enrichment, ICP gates, human approval; ran 100 leads in 48 hours for $4.
launch-factory		Release folder in, launch package out: six assets plus a campaign plan, claims source-traced, human-gated; private for now.
scout-onchain		AI agent on the Zerion API: explainable PnL, technical analysis, DCA windows/alerts, etc.

A harness I have designed and fine-tuned for months: HyperBots. 32 specialist agents, one routing layer, deterministic gates that block a bad ship before it leaves the machine.

This is an operating layer I engineered with context budgeted per seat, routing tested like code, verification gates, and amazing skills, hooks, etc.

→ Router. State the outcome; routing picks the specialist, tie-breaks pre-judged in writing.
→ Mechanical gates. A pre-ship gate blocks every push until a reviewer signs the exact commit. One push human owner.
→ Verification as code. Hash manifests diff source against the installed runtime, trigger eval suites to gate routing changes, and CI catches drift before merge.

The same crew builds the feature, gates the ship, writes the launch copy, and tracks whether AI engines cite it.

An augment is a folder that teaches an AI assistant a whole job. In plain English, it is onboarding for an AI employee, plus guardrails that stop it from cutting corners, zipped so anyone can install it.

Every augment here is three parts:

→ Skills = the SOPs. Step-by-step playbooks the AI follows for each task.
→ Hooks = the guardrails. Automatic checks that block the AI the moment it breaks a rule, instead of trusting it to remember.
→ README = the manual. A stranger installs and runs the whole thing without me in the room.

Three augments, each one shipped real work before it earned a row:

Project	Lang	What it does
worldcup-pari-market		Proof-settled World Cup prediction markets on Solana devnet; demo.
kageb		Private intent pooling for Solana: four real orders in, one aggregate trade out.
grimoire		The Solana Transaction Doctor: paste a failed signature, get the root cause and fix in plain English, live demo.
solana-ship-gate		Pre-deploy safety gate for Solana programs: 4 deterministic checks, blocks unsafe mainnet deploys, MIT licensed.
patronus		Onchain-attested, runtime-scoped credentials for AI agents on Solana DeFi.
worldcup-settlement		An AI agent that bets on live World Cup matches on Solana, autonomously.
meteora-jup-sol-safety-monitor		Read-only Meteora DLMM PnL and safety monitor with Telegram alerts.
solguard		Warns Solana users before they get rugged: scores new pump.fun launches on deployer history, liquidity, mint authority, posts risk live.
superteam-academy		Gamified learning for Solana devs: browser editor, XP, onchain certificates, built for Superteam Brazil.
mermail-defi-navigator		Reads a DeFi email as untrusted data, explains the real yield mechanism, then proposes at most one capped wallet action for Mermail's PayBox to sign.

Project	Lang	What it does
silvia		Food-ordering assistant for older adults: chat instead of learning an app, 50+ completed orders.
safeskill		One-click security audit for your MCP setup: one score, plain English, no CLI required.
wingman		AI Expert Marketplace: install experts (Design, Code, Marketing) into Claude Code.
clawmanship		Reviewable decision before a community skill enters a local bundle.
agenthub		Monad-native AI agent orchestration platform.
hermes-aria-theme		Black and white dashboard theme for the Hermes agent: green for success, red for errors, nothing else.

Project	Lang	What it does
walletbrief-monad		Persistent Monad wallet briefs with human-approved, revoke-only EIP-7702 execution.
yieldpilot		AI copilot for cross-chain yield optimization, built for the Chainlink Convergence Hackathon 2026.

Project	Lang	What it does
beat-claude-engineer-004-analytics-pipeline		Analytics pipeline for the Beat Claude engineer-004 challenge: anomaly detector, test suite, hot-path benchmark.
ecommerce-sales-analysis		E-commerce sales analysis with Python (pandas, matplotlib) and SQL: RFM segmentation, revenue trends, product performance.
solana-narrative-tracker		Spots emerging Solana narratives early: tracks 20+ KOLs and dozens of repos, ships build ideas every fortnight.
defillama-tvl-anomaly-detector		Scans DeFi TVL for what looks wrong: zeroed, -50% in a day, +300% spike, or flat 60 days, catching broken adapters and depegs.

Security researcher, 4 bounties paid. A high-severity Anchor bug in LazyAccount (writeup, PoC, fix, upstream shipped the same fix five weeks later), a fund-lockup bug in Ern Protocol, and two Quantus findings: guardian enrollment without consent in set_high_security, and scheduled transfers below the existential deposit stranding funds with no owner-accessible recovery. Other reports stay unpublished while under review, per program terms.

Also built the Solana Vault Standard Extension for a Superteam bounty.

I write about GTM engineering, all things marketing, web3, proof, evals, and opinions at my blog.

gabeonchain.com · X @gabe_onchain · LinkedIn