A small startup has no traffic. So you never get to see what breaks under scale — and that gap is the biggest obstacle on the way to a global-scale company.
The usual move is to ask “how do I get there?” Charlie Munger goes the other way.
“Invert, always invert.”
Ask “how do I guarantee I stay stuck at this small company forever?” first, and the things not to do become obvious. Stopping those is the roadmap.
TL;DR
| # | Don’t (the trap) | Do (the inversion) |
|---|---|---|
| 1 | Only ever do assigned CRUD | Manufacture the scale you lack |
| 2 | Settle into a familiar framework | Become irreplaceable via CS fundamentals |
| 3 | Sample many tools shallowly | Take one thing to the bottom |
| 4 | Pick jobs by salary alone | Pick by learning density |
| 5 | Postpone English | A little every day (binary filter) |
| 6 | Collect certificates | Leave evidence |
1. You only do what’s assigned
Problem. At a no-traffic company, doing only what you’re told means repeating one year of experience N times. No growth curve on your résumé.
Why it locks you in. Top-tier system-design interviews ultimately ask “what breaks under load, and why?” If you’ve never handled traffic, you have no material to answer with.
Fix. Manufacture the scale your job doesn’t provide.
- Generate load: k6 / Locust / wrk against a personal project
- Data volume: seed millions of rows → profile slow queries → tune indexes/queries → compare p95 before/after
- Inject failure: kill a dependency and watch timeouts, retries, and circuit breakers actually behave
→ If you can’t buy “high-traffic experience,” reproduce it. The numbers (before/after) are the evidence.
2. You settle into a familiar framework
Problem. Proficiency in one framework is a replaceable skill on the market. It doesn’t differentiate you.
Why it locks you in. Google, Meta, and the like filter on data structures & algorithms + system design. They rarely ask which framework you use.
Fix. Invest in the fundamentals that pass the filter: DS & algorithms → OS (concurrency, persistence) → distributed systems. Frameworks are just the surface assembled on top.
3. You sample tools broadly and shallowly
Problem. Ten shallow skills have zero scarcity. Anyone is a substitute → drowned in competition.
Why it locks you in. Career capital comes from the intersection of scarcity × value. Shallow isn’t scarce.
Fix. Take one thing to the bottom. E.g. build a SQLite clone, implement Raft via MIT 6.824, or Build Your Own X. “I implemented this from scratch” is a stronger signal than any certificate.
4. You pick jobs by salary alone
Problem. A high-paying seat with no learning is stagnation. And frequent pivots reset your accumulation.
Why it locks you in. Early-career returns are set by your rate of skill growth (compounding), not salary. Interrupt the compounding and you start over.
Fix. Judge moves by learning density — traffic scale, code-review culture, colleagues stronger than you, difficulty of the problems you’ll own. Salary follows once the skill is there.
5. You postpone English
Problem. In global and remote hiring, English is effectively a binary filter. Pass or fail.
Why it locks you in. Even with enough skill, if you can’t interview in English you’re filtered at the top of the pipeline. Postpone it and it becomes your only rejection reason.
Fix. 30 minutes a day, consistently. Uninterrupted accumulation beats a fancy method (the first rule of compounding).
6. You collect certificates
Problem. Online course certificates and nanodegrees are near-zero trust signals to a senior interviewer.
Why it locks you in. What they trust is verifiable artifacts, not a claim that you “took a course.”
Fix. Leave evidence, not a certificate.
- Open-source merged PRs — code an anonymous senior publicly approved
- A GitHub repo of from-scratch implementations
- A technical blog — evidence of understanding (you can’t write what you don’t understand)
On your résumé, link to these — not to a certificate.
Start this week
- Put load on a personal project with k6; measure p95 before/after on the first bottleneck
- One DS&A problem a day (by pattern)
- Pick one “from scratch” project (DB clone / Raft / etc.)
- Lock in a 30-minute English routine
- Write the whole thing up as one blog post
The point: don’t blame your company for what it doesn’t give you — manufacture the missing environment and leave evidence in numbers and code. That’s the only way to compete with people from high-traffic companies.