I’m a hands-on Fractional Chief AI Officer and former CTO, CTPO and VP Engineering. I help leaders choose where to invest in AI, then work directly with engineering and product teams in their repositories.
I specialise in harness and context engineering. I build the shared instructions, tools, permissions and checks that teams use to develop and review software with Claude Code and Codex.
- Claude Code Ultimate Guide. I wrote it and still maintain it. It includes workflows, templates and exercises teams can use in their own codebases.
- RTK. I was its first contributor, wrote most of its early code and later focused on promoting the project. RTK trims command output so coding agents spend fewer tokens on routine tasks.
I write about AI-assisted engineering: how agents get their context, how teams verify their work, and what the whole workflow costs.
All articles · All practical guides
| Published | Article |
|---|---|
| Oct 14, 2026 | 5 AI code reviewers, 1 repository, and what developers did next |
| Oct 14, 2026 | Turn recurring review comments into tested controls |
| Oct 14, 2026 | A green review is not a merge gate |
| Oct 3, 2026 | Lean and AI: follow the work until it reaches the user |
| Sep 30, 2026 | How to read a token-compression benchmark |
25 articles and guides across four series. Each series page gives the full reading order.
| Series | Parts | Focus |
|---|---|---|
| Engineering the Review Loop | 8 | Local checks, independent review, bounded corrections and merge evidence. |
| The Real Cost of AI | 9 | Task costs, ROI claims, productivity evidence, FinOps and provider dependence. |
| Context Engineering | 6 | Context drift, instruction systems, tool choices and portability. |
| Designing Agent Output | 2 | Flow Lean skills, output styles and checks for their actual behavior. |
- Claude Code Ultimate Guide: a continuously maintained reference for AI-assisted development. Companion pages cover AI FinOps and token-saving tools.
- Claude Cowork Guide: AI workflows for knowledge workers and non-technical users.
- Claude Code security: the attack surface few teams audit: review hooks, connected tools and the permissions they receive.
- Moving the memory observer to a local model: a measured claude-mem case study covering request constraints, timeouts and output quality.
- Context engineering: the L0-to-L5 playbook: choose context controls according to the failure you observe.
- One source configures Claude Code and Codex: shared instructions, immutable releases and skill routing. English · Français.
- From code to merge: my Dev With AI talk, with slides, examples and section recaps. English · Français.
- Dev With AI 2026: conference synthesis and an illustrated report with a TL;DR for each talk. English guide · Guide français · Full report EN · Dossier complet FR.
If you land here from a guide page or an article, starring the repo is the best way to see when something new ships.
- Fractional AI leadership. I work with leadership and product teams on AI priorities, investment decisions and adoption.
- Harness and context engineering. I build shared instructions, tools, permissions, tests and review workflows for coding agents.
- Open source. I maintain the Claude Code Ultimate Guide and build tools such as ctxharness.
- Training and speaking. I run hands-on workshops for teams, contribute to the Dev with AI community and share what I learn in talks and interviews.
Comprehensive resource for AI-assisted development with Claude Code. Documentation-as-product: 473 quiz questions (17 categories), 13 whitepapers (566 pages, FR+EN), 57 recap cards, threat database with 166 catalogued malicious skills/CVEs. 6,000+ stars · 790+ forks · cc.bruniaux.com
CLI proxy that compresses the output of common dev commands (Git, GitHub CLI, Cargo, pnpm, Vitest, Playwright, Docker, Kubernetes) before it reaches the agent. The project reports 60-90% fewer tokens on that output; whole-task studies are collected on cc.bruniaux.com/token-savings. 82,000+ stars · First contributor and main contributor in Q1 2026, then evangelist · rtk-ai.app
Local Rust CLI that picks the model, effort and skills for each phase of a Claude Code or Codex task (plan, execute, verify) and explains why. Limits are checked before scoring, decisions replay to the same receipt, and nothing runs without an approved digest. Experimental: offline routing works; no measured cost gain is claimed yet. Public repository · Build from source with Rust 1.85+ · Source-available (FSL-1.1-ALv2)
Connect Claude, Codex, and other MCP clients to Google Search Console, Bing Webmaster Tools, GA4, CrUX, and public-page analysis through 81 structured tools. Open the bilingual product site · Documentation · GitHub · PyPI
Read-only Rust CLI for macOS to inventory Git worktrees, check commit relationships against selected refs, and inspect local files. A commit can be integrated while untracked or ignored files still need preservation. Backup, restore and removal are unavailable in this MVP. wt-sweep.bruniaux.com · Source-available (FSL-1.1-ALv2)
Map any GitHub repository's stargazers on an interactive world map. Paste a repo URL, get a live geocoded map with clustering, country stats, and an embeddable badge. Next.js + MapLibre + Neon Postgres. starmapper.bruniaux.com
Choose a route based on the outcome you need. Each project appears once under its primary route; the labels show its secondary angles.
Build & Run · Observe & Improve · Secure & Validate · Learn & Adopt · Research, Discover & Grow
flowchart TB
accTitle: AI engineering ecosystem map
accDescr: Five routes connect the ecosystem hub to all 20 projects
hub(["AI engineering ecosystem"])
hub --> build_run["01 · Build & Run"]
build_run --> build_run_projects["CC-Copilot Bridge<br/>Agent Router<br/>Agent Router core<br/>claude-code-plugins<br/>ctxharness<br/>flow-lean<br/>RTK"]
hub --> observe_improve["02 · Observe & Improve"]
observe_improve --> observe_improve_projects["CCBoard<br/>CC-Sessions<br/>cc-skill-usage<br/>Worktree Sweep"]
hub --> secure_validate["03 · Secure & Validate"]
secure_validate --> secure_validate_projects["agentsec-triage<br/>dep-scope"]
hub --> learn_adopt["04 · Learn & Adopt"]
learn_adopt --> learn_adopt_projects["Claude Code Ultimate Guide<br/>Claude Cowork Guide"]
hub --> research_discover_grow["05 · Research, Discover & Grow"]
research_discover_grow --> research_discover_grow_projects["yt-insights<br/>Paper Insights<br/>Search Console MCP<br/>StarMapper<br/>github-roast-tpc"]
classDef hub fill:#f97316,color:#111827,stroke:#fb923c,stroke-width:3px
classDef route fill:#1f2937,color:#f9fafb,stroke:#94a3b8,stroke-width:2px
classDef projects fill:#111827,color:#e5e7eb,stroke:#475569
class hub hub
class build_run,observe_improve,secure_validate,learn_adopt,research_discover_grow route
class build_run_projects,observe_improve_projects,secure_validate_projects,learn_adopt_projects,research_discover_grow_projects projects
Project spotlight: Search Console MCP. Connect Claude, Codex, and other MCP clients to Google Search Console and Bing Webmaster Tools through 81 structured tools. View the repository.
Build, connect, and operate agentic workflows.
| Project | Use it when | Format |
|---|---|---|
| CC-Copilot Bridge · ccbridge.bruniaux.com Provider Routing Local AI Context Engineering |
You need to change the model provider behind a Claude Code workflow. | CLI router |
| Agent Router Model Routing Cost Control Agent Workflows |
Planning, execution and verification should run on different models within a budget, with a decision you can review and replay. Experimental; no measured cost gain is claimed yet. | CLI |
| Agent Router core Model Routing Supply Chain Rust |
You want to build the provider-neutral Rust routing CLI from a fixed snapshot with its distribution inventory, SBOM and licence records. Published on 2026-10-22 and not synchronized with the Agent Router repository. | Rust source |
| claude-code-plugins Agent Extensions Workflow Automation Harness |
You want packaged skills, hooks, agents, and workflows instead of copying files manually. | Plugin collection |
| ctxharness Context Engineering Harness Validation |
You need evidence that CLAUDE.md, AGENTS.md, and related context still match the repository. | CLI |
| flow-lean Context Engineering Prompting Token Efficiency |
You want shorter action-first responses without stacking several overlapping skills. | Claude Code skill |
| RTK · rtk-ai.app Context Engineering Token Efficiency Output Control |
Verbose Git, test, build, or infrastructure output consumes too much context. | CLI proxy |
Inspect sessions, worktrees, and tool usage.
| Project | Use it when | Format |
|---|---|---|
| CCBoard · ccboard.bruniaux.com Observability Session Analytics Monitoring |
You need a complete visual overview of sessions, activity, and analytics. | TUI + Web |
| CC-Sessions Session Search Local-first Knowledge Retrieval |
You need a fast zero-dependency CLI for finding prior decisions or commands. | CLI |
| cc-skill-usage Skill Analytics Local-first Observability |
You need to distinguish real Skill tool calls from prose mentions. | CLI |
| Worktree Sweep · wt-sweep.bruniaux.com Git Worktrees Integration Evidence Read-only |
You need to inventory Git worktrees and inspect their commits and local files before planning cleanup. Read-only macOS MVP; backup, restore and removal are unavailable. | Rust CLI |
Reduce risk and verify code or configuration quality.
| Project | Use it when | Format |
|---|---|---|
| agentsec-triage · cc.bruniaux.com Security Supply Chain Validation |
You need a deterministic read-only scanner for a documented campaign or compromise signal. | Security CLI |
| dep-scope · npm Dependency Analysis Code Quality Validation |
You need to find unused package surface, duplicates, or native alternatives. | CLI + MCP |
Choose a learning route for developers or knowledge workers.
| Project | Use it when | Format |
|---|---|---|
| Claude Code Ultimate Guide · cc.bruniaux.com AI Adoption Context Engineering Security |
You need a maintained reference for architecture, workflows, security, and adoption. | Documentation |
| Claude Cowork Guide · cowork.bruniaux.com AI Adoption Knowledge Work Workflows |
Your work centers on documents, research, analysis, or cross-app tasks rather than software delivery. | Documentation |
Turn source material into knowledge, visibility, or discovery.
| Project | Use it when | Format |
|---|---|---|
| yt-insights · PyPI Research Local Corpus Knowledge Extraction |
You need transcripts, structured insights, local search, and repeatable reports from video sources. | CLI |
| Paper Insights Research Scientific Papers Citations |
You need arXiv metadata search and traceable citations alongside video research. Experimental; human relevance review is pending. | CLI |
| Search Console MCP · search-console.bruniaux.com SEO Analytics Growth |
You need an agent-accessible interface for cross-engine SEO analysis, indexing, and site diagnostics. | MCP server |
| StarMapper · starmapper.bruniaux.com GitHub Discovery Audience Visualization |
You want an interactive view of repository audience and geographic reach. | Web app |
| github-roast-tpc Profile Audit GitHub SEO |
You want README, recruiter-signal, AI-marker, and profile-visibility feedback from a Claude Code plugin. | Claude Code plugin |
| Logo | URL | Description |
|---|---|---|
|
neon.com | Neon sponsors StarMapper and YT Insights. It provides PostgreSQL infrastructure for StarMapper and supports the future hosted data layer planned for YT Insights. |
- 🔜 Nov 5, 2026 🇫🇷 IA : vos développeurs consomment plus de tokens, pour plus de valeur créée ? @ Insitoo Group, with Thomas Pierrain and Vanessa Gossent (live webinar)
- 🔜 Nov 17, 2026 🇫🇷 Je n’ai pas confiance dans mes pull-requests générées @ GenAI Days Nantes (talk)
- 🔜 Dec 15, 2026 🇫🇷 Un agent IA va être jugé @ Power Night, Courbevoie, 18:30 Paris time (roundtable, participating as Lead AI) · Appearance details
- 🔜 🇬🇧 The CTO Playbook, with Adam Horner (podcast, release date to be announced)
- 🔜 🇫🇷 Message à caractère informatique @ Clever Cloud (podcast, release date to be announced)
- 🔜 🇫🇷 Upcoming appearance with Xavier Pestel (Xavki), details to be announced
- Nov 4, 2026 🇫🇷 Contexte Engineering : Dresser ses agents IA (#383) @ IFTTD, with Bruno Soulez (podcast) · Watch on YouTube
- Oct 30, 2026 🇫🇷 DevSecOps compilé: debrief with Bruno Soulez @ IFTTD (podcast, debriefing Tugdual Grall’s original episode)
- Oct 16, 2026 🇫🇷 Coder plus vite avec l’IA, livrer mieux en équipe : du code au merge @ AI-Driven Dev (live)
- Oct 12, 2026 🇫🇷 Coder plus vite avec l’IA, livrer mieux en équipe : du code au merge @ Dev With AI (talk in French) · Article & slides (FR) · Article & slides (EN) · Full conference report (FR) · Full conference report (EN)
- Oct 8, 2026 🇫🇷 Where Cloud Native & AI Intersect - The 2026 Stack @ KubeAuto Day Paris, with Xavier Pestel (Xavki) (panel) · LinkedIn recap
- Sep 2026 🇫🇷 Coder 5x plus vite (#79) @ Tronche de Tech, with Mathieu Sanchez (podcast) · Watch on YouTube
- Jul 7, 2026 🇫🇷 Tokens : le nouveau Cloud Waste @ Generative AI Paris, hosted by WeScale (meetup talk)
- Jun 2026 🇬🇧 Building the StarMapper MCP App @ Alpic - live coding with Frédéric Barthelet
- Jun 2026 🇫🇷 Ce qui t'ouvre les portes des top startups @ GitHub with AI (live)
- Jun 2026 🇫🇷 Tokens : le nouveau cloud waste @ Tech Ready Nantes (talk)
- Jun 2026 🇫🇷 Adoption des agents de code : bonnes pratiques IA @ Tech Ready Nantes (roundtable)
- May 2026 🇫🇷 Context Rot : optimiser son context engineering @ NextLevelRun (live)
- Apr 2026 🇫🇷 Une équipe produit qui shippe avec l'IA with Terry Michel @ Just a Click (podcast)
- Mar 2026 🇫🇷 De CTO avec 30 devs à Solo Engineer boosté à l'IA, GitHub Roast @ The Product Crew (live, source of the github-roast-tpc plugin)
- Feb 2026 🇫🇷 7 mois, 1200 commits, 600 PRs, 50 releases @ Dev with AI (live)
Open a series to see every published part, in reading order.
Engineering the Review Loop · 8 parts
- 1/8 · Coding faster moves the cost to verification
- 2/8 · When an independent reviewer earns its place
- 3/8 · Designing a bounded review loop
- 4/8 · A green review is not a merge gate
- 5/8 · Turn recurring review comments into tested controls
- 6/8 · Operating a multi-reviewer system safely
- 7/8 · 5 AI code reviewers, 1 repository, and what developers did next
- 8/8 · Testing Opus 5.5 as the code-review baseline
The Real Cost of AI · 9 parts
- 1/9 · A $214 AI coding agent rewrite
- 2/9 · Why AI ROI stays invisible, even to the people measuring it
- 3/9 · AI coding productivity: why one gain is not your forecast
- 4/9 · FinOps applied to tokens: who owns the AI bill
- 5/9 · Price per token is incomplete: why cheaper models don't mean cheaper bills
- 6/9 · You cannot depend on a single AI vendor
- 7/9 · Doctolib's cross-verified AI adoption data
- 8/9 · The trillion-dollar backdrop: what markets see that teams don't
- 9/9 · Measuring your own AI usage: three meters, three different answers
Context Engineering · 6 parts
Designing Agent Output · 2 parts
Standalone
- Oct 2026 · Lean and AI: follow the work until it reaches the user
- Sep 2026 · How to read a token-compression benchmark
- Jun 2026 · Claude Is My Second Contributor: What Real Git Stats Show
- May 2026 · Claude Code Under the Hood
- Apr 2026 · Where to Start with Claude Code
- Apr 2026 · From Afterthought to Infrastructure: How AI Config Evolves in a Real Project
- Mar 2026 · AI Velocity Is Bidirectional
- Mar 2026 · Non-Technical to Production in 10 Days with Cursor AI
- Feb 2026 · UVAL: The Protocol I Built to Stop Accepting Code I Don't Understand
- Feb 2026 · From VP Engineering to Solo Builder: My Honest Take
As Founding AI Engineer at Méthode Aristote from July 2025 to October 2026, I built the initial EdTech codebase from scratch, developed the team’s AI harness and trained product builders to contribute to production. I shared the review workflow in From code to merge. Our collaboration continued on a freelance basis after the salaried role ended.
AI since 2017: NLP and predictive ML in production at Q°emotion and eXplain (semantic search, sentiment scoring, ~30M articles processed). Generative and agentic AI daily since 2025.
12+ years: Developer → Team Lead → EM → VP Engineering → CTO → back to hands-on by choice.
Why hands-on? After a decade leading teams (4-150 people, managed up to 30 engineers), I deliberately stepped back to sharpen my technical edge. Leadership is more effective when grounded in current engineering reality. In the first 3 months of 2026, that bet produced 10+ open-source projects, 5,000+ stars on my own, plus core contributions to RTK (82,000+ stars).
Leadership experience:
- Built and scaled engineering teams from scratch (early-stage → Series A → scale-up)
- Shaped engineering culture, processes, hiring, mentorship
- Led across SaaS, E-commerce, Data, EdTech domains
Tech:
- Stack: TypeScript (Next.js 14+, Nest.js, tRPC), Rust, Ruby (Rails), Python, SQL/NoSQL
- AI: Claude, GPT, MCP server development, agentic workflows. NLP/ML since 2017 (Elasticsearch, spaCy, Gensim)
- Infra: AWS, GCP, Vercel, Docker, Kubernetes, serverless
- Domains: Fullstack (backend preference), product development, technical architecture
Formation: Computer Engineering @ UTT Languages: French, English
- 🌐 florian.bruniaux.com - Portfolio
- ✍️ florian.bruniaux.com/blog - Blog on AI-assisted engineering, updated most weeks
- 📂 cc.bruniaux.com - Claude Code Ultimate Guide, continuously updated
- 🎤 florian.bruniaux.com/media - Talks, podcasts & live streams
- 🌍 devw.ai - Dev with AI, French-speaking AI community
- 💼 cowork.bruniaux.com - Claude Cowork Guide
- 🗺️ starmapper.bruniaux.com - StarMapper
- 📊 ccboard.bruniaux.com - CCBoard
- 🌉 ccbridge.bruniaux.com - CC-Copilot Bridge
- 🦀 rtk-ai.app - RTK - Rust Token Killer