Archived 2026-10-01. Superseded by the three-course series — SWE 101 → SWE 102 → SWE 103. Kept for the reasoning it records; it is no longer the plan.

Weeks 1–28. The track that produces the offer. At 10–15 h/week this is roughly 8–10 of those hours.

The one rule: this track wins every conflict with Track B. When a week collapses, DSA and applications survive.


The weekly rotation

JAMB shape — one thing per day, fixed, not everything every day.

DayFocusHours
MonDSA — one pattern, 2–3 problems1.5
TueFoundation topic (Track B)1.5
WedDSA — same pattern, cold1.5
ThuBuild — flagship1.5
FriApplications + outreach1
SatBuild block3–4
SunReconstruct & explain — close everything, teach the week from memory1

≈ 11–12 hours. Sunday is not optional; it’s step 3 of the loop and it’s the one that gets skipped.

Friday is non-negotiable from week 7. An hour a week of applying, forever, beats a heroic month of it later.


Phase 0 — Audit and artifacts · weeks 1–2

Before any studying. These are the things that are worth more per hour than anything else in the plan, because they’re already 90% done and just aren’t visible.

  • Pick the flagship. One project a stranger can click. Recommendation: harden nextvibe or my-applicant rather than starting fresh — a new build costs 8 weeks to reach where these already are.
  • Publish the three blog drafts sitting in blog-drafts/ — four-bugs-that-shipped, socket-race-condition, reading-a-datasheet. These are exactly the posts that get a remote junior noticed: specific, debugged, real. They are finished. Publishing them is an evening.
  • CV — one page, targeted at TS full-stack + AI product. Projects, not coursework.
  • GitHub — pin 4 repos, real READMEs with a screenshot/GIF and a live URL at the top. A repo with no README is invisible.
  • LinkedIn/X — headline says what you build, not what you’re studying.
  • Make the Quartz site discoverable — link it from GitHub, CV, LinkedIn. 1,150 notes published is a genuine differentiator and right now nobody hiring can find it.
  • DSA baseline — sit 3 unseen mediums, 45 min each, cold. Score honestly. This calibrates everything.

Phase 1 — DSA, continuously · weeks 1–28

Not a block. Never dropped. Two sessions a week for the whole track, because a skill crammed in weeks 8–12 is gone by week 30.

Method is yours and doesn’t change:

Study 1–2 worked solutions → extract what makes it that kind of problem → close them → solve new ones cold. Never copy a third.

Work through the 15 patterns already written here. One pattern per week, roughly:

arrays/hashing → two pointers → sliding window → binary search → stack → linked list → trees/BFS/DFS → heap → backtracking → graphs → intervals → greedy → DP (2 weeks) → tries → bit manipulation

Target: 150 problems, ~60% medium. Log every one in the notebook with pattern, whether it was cold, and time taken. The log is the visible progress — this is the board.

Track the tell: if you can’t restate the pattern on a blank page before starting a problem, you’re pattern-matching, not pattern-knowing.

Phase 2 — The flagship · weeks 3–12

One project, shipped properly. Not five half-projects.

What it has to have, in order:

  1. Deployed, with a URL a stranger can open — no local-only
  2. A README that opens with what it does and a GIF
  3. Real tests, running in CI
  4. An LLM feature with an evals harness ← the differentiator
  5. An architecture note explaining one hard trade-off you made
  6. Observability — you can answer “is it up, and is it fast”

Item 4 is the whole point. A golden set, a scoring function, a regression run on every prompt change, and numbers you can quote. Almost nobody applying to junior AI roles has this. The note is already written — this is the reps for it.

Milestones (each independently shippable, per the build-guide rule):

  • w3–4 — scope, deploy the skeleton to production on day one, CI green
  • w5–6 — the core feature end to end
  • w7–8 — the AI feature working
  • w9–10 — evals: golden set, scorer, CI regression
  • w11 — hardening: errors, rate limits, cost/latency numbers
  • w12 — README, architecture note, demo GIF, ship the write-up

Phase 3 — Apply, from week 7 · weeks 7–28

Start before it feels ready. This is the instruction most likely to be ignored, and the one that costs the most.

Applying is not the reward for finishing. It’s a diagnostic you cannot get any other way: rejections tell you which gap is actually binding, and that’s information that changes what you study. Waiting until the portfolio is perfect means finding out in month 9 what you could have found out in month 2.

  • Weeks 7–12: 5/week. Learning the machinery — CVs, cover notes, take-homes, what gets replies.
  • Weeks 13–28: 10/week. Volume. Target 100 by week 20.
  • Fridays, every week, one hour. Non-negotiable.

Where remote juniors actually get hired: AI-product startups (small, portfolio-led, often skip LeetCode for a take-home), YC company job boards, Wellfound, remote-first job boards, and — highest hit rate of all — being visible: writing, shipping in public, being useful in a Discord/GitHub for a tool you actually use.

Keep a spreadsheet: company, date, source, stage, outcome. Track reply rate. Below ~5% after 30 applications, the CV or the targeting is wrong, not the market.

Phase 4 — Interview reps · weeks 12–28

Runs alongside applications, not after.

  • Coding rounds — the DSA log is the prep; add timed mock sessions from week 12
  • System design — the round, then one design per fortnight from Track B’s list, out loud, whiteboard, 45 minutes
  • The project story — nextvibe’s story bank is the model. Be able to tell the flagship in 2 minutes, 5 minutes, and 20 minutes of depth.
  • Behavioural — 6 stories, STAR, from real projects. Remote roles probe async communication hard.
  • The banks already exist — 13 domains. Cover the answer, say it out loud, then compare. Recognition is not knowledge.

After every real interview, in the notebook: what was asked, what I fumbled, what I’ll fix. This is the highest-value page in the book.


What “on track” looks like

WeekShould be true
2Blog posts live, CV done, GitHub pinned, DSA baseline scored
7Flagship deployed and CI green · first applications out
12Flagship has evals · 30 DSA problems · applications weekly
20100 applications · 90 problems · interviews happening
28150 problems · loops in progress

If week 7 arrives and nothing has been applied to, that’s the failure mode this whole restructure exists to prevent. Not the DSA count. The applications.

↑ Back to SWE 101 — Archive