Phase 7 — Production & Mastery (M12–M13)

This is the payoff phase. Everything from M1 to M11 was buildup; M12 and M13 are where you become what the pitch says. By the end of this phase you have merged a non-trivial patch into a major C open-source project, you can talk about how the pros write C without hedging, and you have at least one war-story-worthy public artifact — a blog post, a talk, or a bug hunt writeup. The goal is not “more knowledge.” It is credible signal.

The M13 pitch, again, as a mantra: “I write C that ships in production — from SIMD-accelerated ML inference kernels to multi-threaded network services — and I can debug someone else’s segfault at 2 AM without flinching.” Everything below is graded against that sentence.

The two-month arc

The first month (M12) is about ingesting production C at scale and preparing your first PR. The second month (M13) is about landing that PR, writing the artifact, and mapping your next role. Both months are 10–15 h/week. If you skip the PR gate, you did not finish this roadmap.

Month	Weeks	Focus	Deliverable
M12	W1	Read sqlite3.c amalgamation & first 10 chapters of doc/architecture.md	Repo notes in notes/sqlite/
M12	W2	Read Valkey src/server.c, networking.c, t_string.c, ae.c event loop	Repo notes in notes/valkey/
M12	W3	Static analysis + fuzzing your gemm_bench and quantize_ext repos	clang-tidy clean + one libFuzzer harness
M12	W4	Pick issue, comment on it, open draft PR	Draft PR link
M13	W1–2	PR review loop: iterate on maintainer feedback	Merged PR (the hard gate)
M13	W3	Write 2–3k-word technical post about the journey	Published post URL
M13	W4	study prep: role map, resume rewrite, 3 mock studies	Role list, updated resume

File map

#	File	What you get
01	01_reading_production_code.md	The four canonical codebases and the order to read them
02	02_static_analysis_and_ci.md	clang-tidy, cppcheck, oss-fuzz, GitHub Actions matrix
03	03_secure_c_and_cve_literacy.md	CWE top-25 for C, real 2024–2026 CVEs, how bugs become CVEs
04	04_portability_and_abi.md	Why “just recompile” fails, symbol versioning, semver for C
05	05_contributing_to_oss_c.md	How to find, scope, and land your first C PR
06	06_the_c_expert_reading_list.md	15 talks and papers every serious C engineer has consumed
07	07_interview_and_role_map.md	Who hires heavy C in 2026 (global + India), how to decode the JDs
08	08_war_stories_the_hard_lessons.md	Eight production disasters with sources — the folklore
09	projects.md	The final two deliverables: merged PR + public artifact

Exit criteria — you cannot skip these

You are done with Phase 7 (and this roadmap) when every one of these is true. No self-grading loopholes.

• You have a merged pull request in one of: Valkey, curl, llama.cpp, redis, PostgreSQL, ClickHouse, or an equivalently-serious C project. The PR must touch functional code (not a typo fix). Link committed to your 99_pre_mortem/wins.md.

• You have published a technical writeup of at least 2,000 words about one deep C concept from your own journey — cache blocking, GGML quantization, a CVE analysis, or your PR’s debugging story. Public URL committed.

• You have run clang-tidy, cppcheck, and one fuzzer (libFuzzer or AFL++) against your own gemm_bench or quantize_ext repo, and fixed every finding you can defend fixing.

• You can name three currently-active maintainers each in Valkey, llama.cpp, and curl, and describe their contribution style from memory.

• You can walk through the SQLite architecture (VDBE → B-tree → Pager → OS interface) on a whiteboard without notes.

• You have assessed at (or been rejected by) at least one company from 07_interview_and_role_map.md and written the retro.

• 99_pre_mortem/final_retro.md exists, with the honest audit: what you can now do, what you still can’t, what you’d re-do if you started over.

Time budget (M12–M13 combined ≈ 96 h)

Bucket	Hours	Notes
Reading production code	24	~6 h/week for four weeks
Static analysis + fuzzing	8	Your own repos first
PR work (find → land)	30	Includes review-cycle wait time that you fill with reading
Writeup	12	Draft, revise, revise, publish
study prep	12	Role research + resume + 3 mocks
Slack + retros	10	Because life happens

What “done” feels like

You should be able to open a stranger’s C repository — say, a small database or a network daemon — clone it, build it, find one bug or missing feature, and produce a defensible patch inside a weekend. Not “understand the code eventually.” Not “read a tutorial about it.” Ship the patch. That is the tell.

That instinct is the entire point of thirteen months. You built it, brother.

---

Return to ../README.md · Next: 01_reading_production_code.md