AVAILABLE FOR FULL-TIME ROLES OPEN TO FREELANCE CONTRACTS AI FULLSTACK ENGINEER · AGENTIC AI AGENT-READY · HUMANS WELCOME AI AGENTS · LLM TOOLING · EVALS NEXT.JS · TYPESCRIPT · REACT BASED IN LISBON · REMOTE WORLDWIDE
01 / Case study — autonomous research

ECDSA.fail, taken by a machine that runs itself.

An autonomous multi-agent research harness that attacks the ECDSA.fail benchmark around the clock. Nine specialist LLM agent roles routed across seven-plus providers, dispatched through the filesystem with lane isolation, held behind spend gates and human approvals — and wired to a public telemetry dashboard so every token is accounted for.

Role
AI engineering lead
Org
Quantum
Period
2025 — 2026
Result
#1 on ECDSA.fail
9 Agent roles Specialist LLM roles, each with its own prompt, lane, and model routing
7+ Providers Claude · Kimi · DeepSeek · Gemini · GLM · Codex · Amp
30m Loop period launchd discovery → validate → submit, around the clock
#1 Result Top of ECDSA.fail — research contributor on the publication
01.1 / The harness

A role→model router assigns every specialist role a primary model and a health-probed fallback chain across providers. When a provider degrades, the chain degrades gracefully — the campaign doesn't.

a fig. 01 — routing fabric health-probed fallback chains
  discover ──▶ route(role) ──▶ ┌ claude ──┐
      ▲                        ├ kimi      ├─▶ dispatch ──▶ validate ──▶ submit
      │                        ├ deepseek  │        ▲
      └────────── 30 min ◀─────┤ gemini    ├────────┘
                               ├ glm       │   worker envelope
                               ├ codex     │   (strict JSON)
                               └ amp      ─┘
        probe: unhealthy provider ⇒ next in chain
b fig. 02 — dispatch discipline fail-closed by default
  • file-based dispatch work units move through the filesystem — no broker, no hidden state
  • lane isolation + PID locks agents can't collide; a stale lock is evidence, not a hazard
  • fail-closed adapters any malformed provider response closes the lane instead of corrupting the run
  • worker-envelope validation strict JSON schema on every worker result before it can touch state
  • tamper detection git-status snapshots flag any drift between what ran and what's on disk
01.2 / Safety & telemetry
c fig. 03 — spend gate hard ceiling

A $200 spend gate caps autonomous burn; GPU-heavy steps require human approval before they launch. The harness is autonomous inside the envelope, obedient at its edges.

d fig. 04 — public telemetry every token accounted
  • token spend per provider, live
  • multi-objective frontier candidate quality vs. cost, plotted
  • falsifier queue candidates scheduled for attempted refutation
01.3 / Outcome
#1

The campaign took the top spot on ECDSA.fail — and the harness itself became part of the result: research contributor on the resulting publication.