01 · Roasts
The 10x Prompter's Portfolio
You've shipped 6 AI agent orchestrators in 6 weeks (openfleet → flt → harness → agentelo → claudecord → hone). At some point 'abstracting CLI adapters' stops being a product and starts being a coping mechanism.
Sprint God, Star Peasant
2,862 commits this year, 165 PRs, live SaaS products with Stripe billing — and still only 5 followers. You're building in a soundproof room and calling it a launch.
Architecture Docs Enjoyer
flt alone ships DESIGN.md, STATUS.md, PROJECT_STATE.md, docs/adapters.md, docs/architecture.md, docs/workflows.md. Nobody needs 6 documentation files for an 11-day-old repo. The README was fine.
Burst Builder
openfleet: 6 days. harness: 4 days. claudecord: 4 days. hone-a-drone: 1 day. You don't iterate on projects, you ritually abandon them the moment the architecture is 'clean enough'.
Self-Described 10x Prompter
Bio says '10x prompter'. The 99% solo commit rate confirms it. At least give Claude Code a co-author credit — it's clearly doing half the heavy lifting here.
Built using
Zoral
Shadows one worker for a week, then takes over their job with zero extra setup. Behaves exactly like the original.
zoral.ai
02 · Category breakdown
- Impact25% weight68C
- Consistency20% weight95S
- Quality20% weight77B
- Depth15% weight65C
- Breadth10% weight65C
- Community10% weight50D
03 · Stats
365-day commit heatmap
122 active days
Language distribution
- TypeScript72%
- JavaScript12%
- Python5%
- Lua5%
- Shell2%
- C2%
- Other2%
04 · Numbers
Owned repos
non-fork
14
Commits
last 12 months
2,862
Followers
5
Joined GitHub
Nov 2021
05 · Top repos
twaldin /
trade-up-bot
CS2 trade-up contract analyzer with live web app (tradeupbot.app), TypeScript, PostgreSQL, multi-source market pricing (CSFloat/DMarket/Skinport), probabilistic EV calculation, and claim/verify system. ~56 server + 38 frontend files, 20K LOC, well-tested.
twaldin /
agentelo
Production-grade ELO benchmarking platform for AI coding agents on real GitHub bugs. Typed TypeScript/Node backend with Bradley-Terry rating math, SQLite persistence, multi-harness orchestration, and Next.js frontend. Real market use (148 agents ranked, 5+ harnesses supported).
twaldin /
tim.waldin.net
Polished indie terminal portfolio SaaS with Docker-isolated containers, Socket.IO streaming, Next.js+React frontend, and hardened security—functional production tool with comprehensive docs, typed code, and multi-file architecture.
twaldin /
flt
TypeScript agent orchestration harness for coordinating multi-LLM workflows. Fully typed, tested, documented, ships with CI/CD. Recent sprint (~11 days old) with 30 commits across structured architecture (adapters, state, TUI, workflows).
twaldin /
harness
Cleanly typed Python+TypeScript library abstracting 8 AI CLI agents (claude-code, opencode, codex, gemini, aider, swe-agent, qwen, continue-cli) into a unified RunSpec/RunResult interface; ships with tests, CI, comprehensive SPEC.md + ARCHITECTURE.md, and fixture-driven cross-language verification.
twaldin /
openfleet
Harness-agnostic AI agent orchestrator supporting Claude Code, OpenCode, OpenClaw, Codex. Well-documented portfolio project with modular core, multi-harness adapters, test fixtures, and real skills/instructions system. No type annotations (plain JS), but structured, shipped, and non-trivial.
twaldin /
hone
Specialized Python CLI optimizer composing GEPA + agentic coding CLIs (Claude Code, Gemini) as mutators against user graders. v0.3 adds multi-file targeting with scheduler, ACE observer, persistent run tracking. Well-structured typed codebase with tests; thin on-disk footprint (68 KB) reflects focused scope.
twaldin /
claudecord
TypeScript MCP server bridging Discord and Claude Code agents with message routing, channel lifecycle management, and permission prompts. Early-stage plugin (3 stars, 4 days old) with structured architecture and comprehensive tests.
twaldin /
hone-a-drone
Brand-new drone racing controller evolved via hone optimization against lsy_drone_racing sim. Demonstrates solid module architecture and clear evolution targets, but 1-day-old with minimal external adoption. Active but experimental.
twaldin /
studyspot
TypeScript Next.js + Cloudflare study platform with RAG-powered AI tutoring. Typed + documented (README describes architecture), but no tests, CI, or source files visible; young repo (9mo) with modest commit activity.
twaldin /
tmux-orchestrator
Brand-new (2 days old) tmux-based Claude Code plugin with clear product vision but minimal codebase. Has README and basic shell scripts but no tests, CI, or production validation. Single-week sprint tier project.
twaldin /
twaldin
Personal profile README with links to other projects (agentelo, trade-up-bot, flt). 10KB repo, 7 of last 30 commits, created March 2026. No code, tests, CI, or license.
06 · Timeline
- Nov 17, 2021Joined GitHub
- Jul 3, 2025Created studyspot
- Aug 15, 2025Created tim.waldin.net — personal website but its bash
- Mar 16, 2026Created trade-up-bot — Finds profitable trade-up contracts using real marketplace listings in CS2
- Mar 19, 2026Created twaldin — Profile README
- Mar 28, 2026Created claudecord — Discord Server architecture for multi-agent orchestration and automation
- Mar 30, 2026Created tmux-orchestrator — Better agent teams for claude code: manage work across multiple projects and worktrees, no dependencies, just bash
- Apr 1, 2026Created openfleet — Manage a fleet of openclaw agents using your provider subscription through their native cli
- Apr 5, 2026Created agentelo — public ranking system for ai agents
- Apr 12, 2026Created flt — cli-first harness agnostic agent orchestration tool
- Apr 18, 2026Created hone — CLI text optimizer built on GEPA. Uses Agentic Coding CLI's as mutator and observer -- no api keys required
- Apr 18, 2026Created harness — Unified Python interface for invoking AI coding-agent CLIs (claude-code, opencode, codex, gemini, aider, swe-agent) as subprocesses.
- Apr 20, 2026Created hone-a-drone — hone-driven evolution of drone racing controllers against lsy_drone_racing sim. +33% aggregate gain, +270% on L2 from a single budget-100 run.
- Apr 23, 2026Most recent push to flt
07 · Compare
08 · Rubric
How this score was produced
Overall = Σ (category × weight) + gentle top-end curve
Tier thresholds
▸ How the pipeline works
- 01Scrape.Pull every non-fork repo pushed in the last 90 days, plus your contribution calendar, followers, and language byte counts — straight from GitHub's REST & GraphQL APIs.
- 02Triage.A small model reads every repo's file tree + README and picks the 20 files per repo that actually reveal how you code.
- 03Grade each repo. All repos run in parallel through a fast scoring model that reads the picked files and rates each one independently on Impact, Quality, and Depth — with evidence citations.
- 04Aggregate. A larger reasoning model combines the per-repo scores with server-computed stats (heatmap, commit cadence, language entropy, follower count) to produce the 6-dimension profile score + roasts.
- 05Correct.Deterministic server-side checks enforce anchor-scale floors (e.g. a profile with 2,000+ public commits can't score 30 Consistency) and recompute the final verdict.
~90 seconds per profile, ~$0.25 in compute. Total of ~240 files read across your top-12 repos. One rating per GitHub account per day.
▸ Data sources & caveats
- Heatmap & commit totals: GitHub GraphQL
contributionsCollection— covers the last 365 days, includes private repos when the user has opted in (default). - Language %: byte totals across the top 30 owned non-fork repos.
- Curve: a small upward nudge centered on raw score ≈ 70, capping at 100. Prevents specialists from being unfairly penalised for narrow breadth.
- Anchor corrections: when server-measured signals (e.g. privateWorkLikely, multiRepoVolume, follower count) mandate a minimum category score, the aggregation step enforces it. These are signal-conditional, not identity-based floors.