Today's Edge
Your one thing: kaggle-automation — Python FastAPI that shards GPU jobs across Kaggle accounts, 12h monitor. It's the only project you can defend end-to-end in Python right now. Don't chase the Rust tokenizer until you can talk Python for 10 min without blanking. Every task today ladders to this.
LeetCode — protected tests
AI code — spot subtle bugs
Portfolio — why this stack?
ML from scratch
Before you apply anywhere today — classify the JD (A/B/C/D) then copy-paste to recruiter:
Be T-shaped
- Product thinking: write 1 user story for kaggle (“As a user I want to shard 10GB”)
- Calm in chaos: 25-min struggle cap then read editorial, no spiral
- Communication: 10 min explain your stack aloud, no notes
Brutal truth for today
You graduated 4 months ago. You have 4 years of screenshots and 34 repos but only 2 LeetCode problems actually solved. Collecting advice is not progress. Today you do 2 heap problems + 1 real commit + 3 tailored apps — nothing else counts.
You start at 13:22. Morning 07:00–13:05 is marked skipped — don't guilt-spiral. Do the remaining blocks only. MVP is still 2 heap + 1 commit + 1 app — heap missed today: do 1 heap at 14:00 inside Project 2 time, or at 18:15 if you run out.
07:00–13:05 — SKIPPED (late start) — tap to expand
Why? roadmap.sh/backend expects DB depth + caching + auth before queues/K8s — you have 0 Redis evidence. These 10m close the gap.
Verified resources (live Sep 4): yt-ssf (Greg Hogg Heaps · Docker 2026 · REST Hayk) · aman.ai-search (Docker run · Bias/Variance) · archive: Bandit L0–5 + SQL roadmap + X-Y-Z resume notes.
Minimum viable day
2 heap problems + 1 app + 30m docker — never zero. Streak > perfection.
If you only do 3 things tomorrow
- Ship the Dockerfile —
docker build .passes - Finish Kth Largest + Top K Frequent without looking
- Send 1 DM with that recruiter line + your GitHub link asking for feedback, not “any openings?”
Files you opened today are not progress. Shipped things are. Every 25-year-old who got hired after 4 months did these boring reps when no one was watching.