Nine things I wish I knew earlier
Short, practical lessons from building and shipping software, AI systems and products.
Software engineering
Boring tech, sharp product
Spend your novelty budget on what makes the product different. Postgres, a queue and a cache solve most problems.
Every bug fix ships with a test
The test that would have caught it is the only proof it stays fixed.
Assume every network call fails
Timeouts, retries with backoff and jitter, and idempotency keys. No infinite defaults.
AI / ML engineering
Write evals before prompts
A 50-example golden set from real inputs turns "feels better" into a number.
Retrieval quality caps RAG quality
Measure recall@k separately from answer quality before touching the model.
Beat the baseline first
If a heuristic or logistic regression is close, the complex model is not worth shipping.
Entrepreneurship
Sell before you build
A landing page, a waitlist or a pre-sale tells you more than a month of code.
Distribution is the product
Pick one or two channels and go deep for 90 days before adding more.
Charge earlier than feels comfortable
Free users give opinions. Paying users give the truth.
Want the full playbooks?
Two free PDFs. Pick one or grab both.
Free PDF
Engineering Tips
30+ field-tested rules for building reliable software and AI systems.
- –Design, testing and production habits
- –LLM, RAG and agent engineering
- –ML pitfalls that ship bad models
- –Launch and production-readiness checklists
Free PDF
Founder Tips
20+ practical lessons for validating, launching and growing your own product.
- –Validate before you build
- –Distribution and SEO that compound
- –Pricing and retention basics
- –A 90-day launch checklist