Latest learning
7 min read
Skills should declare capabilities, not model names
Why portable agent skills need provider-neutral intelligence and reasoning requirements, and how Fleet is testing the idea.
Read the learningSoftware as a specialized service
A living index of things I build.
26 maintained products, with their real domains, roadmaps, release history, and someone accountable for every one.
Explore the workBuilt in public. Kept current.
- 01 CodeVetter AI code review platform — desktop-first, works offline. codevetter.com
- 02 PostTrainLLM A local factory for training, evaluating, and running specialist language models. posttrainllm.com
- 03 HeyPace Local-only macOS voice agent that can understand what is on your screen. heypace.app
- 04 High Signal Evidence-backed daily intelligence across technology, startups, finance, and public markets. highsignal.app
A cumulative studio measure
TOKENS SPENT
FOR THE WORLD
Model tokens used while building and serving SaaS Maker products—counted from verified product records, never estimated.
A transparent computation ledger, not a proxy for people reached or social impact.
165,444,915,616 tokens spent for the world
VERIFIED SNAPSHOT · AUG 9, 2026 LAST UPDATED AT · AUG 9, 2026, 12:20 UTC
- Latest day
- 65,896,278 Aug 9, 2026
- Projects used
- Worldwide Across borders and time zones
- Projects contributing
- 1 CodeVetter starts the count
Verified CodeVetter usage baseline. Additional products join the cumulative total through authoritative daily seeds.
Projects used worldwide.
The maintained catalog
26 products.
One accountable maker.
Each product keeps its own domain, roadmap, changelog, and release boundary. Nothing here is a concept render.
Skip to latest learningProducts & research
- App Health Privacy-first endpoint health for Node, Go, and OpenTelemetry services.
- Free AI OpenAI-compatible gateway across free-tier model providers.
- Knowledge Base Private agent search over specialized corpora with ranked citations, provenance, and schema-aware retrieval.
- Research Papers Academic paper discovery and a structured research data asset.
- Starboard Organize and semantically search your GitHub stars.
Maker tools
Personal utilities
- Anime List Anime and manga discovery with multi-axis filtering and personal watchlists.
- Calorie A private, local-first food, water, and weight journal with transparent timing guidance.
- Chess Coach Browser chess against Stockfish with optional AI move coaching.
- Email Manager A private Gmail workspace for local semantic search, sender insights, and explicit unsubscribe workflows.
- Karte An AI link-in-bio that turns a profile into a conversation.
- LoopTV A lean-back, TV-style random video player for curated channels.
- Memory Map Turn a ChatGPT export into a private, browser-computed map of recurring themes, facts, and conversations.
- Motion Use your body as the controller for an iPhone-hosted game that can mirror to a larger screen.
- Reader Capture, annotate, revisit, and discuss saved reading.
- RolePatch AI-assisted resume tailoring, role research, and interview preparation.
- Sarthak Agrawal Personal portfolio of Sarthak Agrawal — AI infrastructure and product engineer.
- Setline A mobile-first workout execution tracker for following user-authored plans, recording sets, and timing rest offline.
- Significant Hobbies Plan meaningful hobbies, side quests, rituals, and life journeys.
- SWE Interview Prep A learning OS for software-engineering interview practice.
- What It Takes to Win Explore 2,585 documented early-breakthrough paths without pretending success follows a formula.
Public repository history
Past work, kept legible.
10 public repositories preserved for source and attribution, without pretending they are still maintained.
- AliveVille Persistent AI world and multi-agent simulation experiments.
- Companion Robot Earlier companion robotics and embodied-agent exploration.
- EverythingRated A general-purpose rating and comparison experiment.
- Forecast Lab Forecasting experiments and evaluation tooling.
- Materia Evidence-graded remedy reference organized by body system.
- Open Historia Interactive historical storytelling and world simulation.
- Protein Index Searchable protein and nutrition reference.
- Today Little Log A small daily logging experiment later folded into other work.
- TrueHire Earlier resume and hiring workflow superseded by RolePatch.
- Web Playables Small browser-playable game experiments.
Open-source package
@saas-maker/feedback
A small React widget for bugs, feature requests, screenshots, and page-specific feedback. Your application owns submission and storage.
Selective commissions
Bring me a problem worth building.
I mostly build my own products. Occasionally, I take on a focused external build with a clear problem, a serious owner, and a real budget.