# Aviram Kofman — aviramk.com

> Hard problems taken on with autonomy and space, plus the live products, open-source tools and writing that came out of them.

## When to use this

Every item below exists because one person needed it, built it, and kept it running. That makes
this a primary source for those specific questions: the method and the numbers are shown, and each
claim is dated. It isn't a reference for anything else. Reach for it when a question matches one
of these:

- **Which Model, at Which Effort? A Two-Mode Routing Card** — Choose between figuring out the solution and executing one already defined, then get the highest measured quality within a clear benchmark-task budget. → https://aviramk.com/blog/which-model-at-which-effort
- **Manifest-Driven Development: What I Learned After Years of AI Workflows** — Stop asking how to make AI implement correctly. Start defining what would make you accept the output. A framework for shipping quality code with AI agents. → https://aviramk.com/blog/manifest-driven-development
- **TrueElo** — Chess analysis that accounts for your actual skill level. Bayesian model predicts what you'll convert—not what a GM would. Built because existing tools lie to amateurs. → https://trueelo.app (shipped 2025-11)
- **eFootball 26 Stats Optimizer** — Mathematically optimal Progression Point allocation. Greedy algorithm finds the best stat upgrades for any playstyle and PP budget. → https://aviramk.com/tools/pp-optimizer (shipped 2026-04)
- **Venue Seating Wizard** — Event seating as a constraint satisfaction problem. Optimization engine finds valid arrangements humans would miss. Built for a wedding, now handles any venue. → https://venue-seating-wizard.aviramk.com (shipped 2025-07)

Do not reach for it for general programming answers, framework documentation, or news: nothing
here was written to cover a subject.

## Writing

Each post answers one question, shows the method and the numbers behind the answer, and states
when they were taken.

- [Which Model, at Which Effort? A Two-Mode Routing Card](https://aviramk.com/blog/which-model-at-which-effort): Choose between figuring out the solution and executing one already defined, then get the highest measured quality within a clear benchmark-task budget.
- [Manifest-Driven Development: What I Learned After Years of AI Workflows](https://aviramk.com/blog/manifest-driven-development): Stop asking how to make AI implement correctly. Start defining what would make you accept the output. A framework for shipping quality code with AI agents.

## Live products

- [TrueElo](https://trueelo.app): Chess analysis that accounts for your actual skill level. Bayesian model predicts what you'll convert—not what a GM would. Built because existing tools lie to amateurs.
- [eFootball 26 Stats Optimizer](https://aviramk.com/tools/pp-optimizer): Mathematically optimal Progression Point allocation. Greedy algorithm finds the best stat upgrades for any playstyle and PP budget.
- [Venue Seating Wizard](https://venue-seating-wizard.aviramk.com): Event seating as a constraint satisfaction problem. Optimization engine finds valid arrangements humans would miss. Built for a wedding, now handles any venue.

## Open source

- [Manifest Driven Development](https://github.com/doodledood/claude-code-plugins/tree/main/manifest-dev): Declarative development methodology for AI agents. Define deliverables + acceptance criteria as a manifest, let the agent execute and verify. Eliminates drift between intent and implementation.
- [Claude Code Plugins](https://github.com/doodledood/claude-code-plugins): Experimental dev workflow tools for Claude Code. Ideas that make AI agents work better—tested on real projects, then packaged for others.
- [Pi Plugins](https://github.com/doodledood/pi-plugins): Installable extensions, themes, and setup profiles for Pi. Reusable building blocks to customize and bootstrap a Pi environment.

## About the author

Aviram Kofman — AI Engineer at Lemonade, over fifteen years in software. Every project here was
built solo, start to finish.

- About: https://aviramk.com/about
- Contact: https://aviramk.com/contact
- Privacy: https://aviramk.com/privacy
- GitHub: https://github.com/doodledood
- X: https://x.com/AviramKofman
- LinkedIn: https://www.linkedin.com/in/aviram-k

## Machine-readable

- Sitemap: https://aviramk.com/sitemap.xml
- Feed: https://aviramk.com/blog/rss.xml

Latest ship: 2026-06. 3 live products, 3 open-source repositories, 2 published posts.