Build vs Buy: Software Systems at Jurassic Park
summary
Retrospective of Jurassic Park engineering details why building custom security, monitoring, telephony, video surveillance and door locks costs far more than two weeks of coding: hidden edge cases (no manual door override), ongoing maintenance, single-developer risk called the Dilophosaurus factor, and a cost model that compares a $120k engineer to a vendor at $3,000/month. Vendor risk appears too when third-party Raptor collapses after a left-pad dependency, and real-world stats show over 50% of large implementations still fail. Five decision questions: impact of failure, novelty, bias, opportunity cost, and sustainability.
AI-written summary · about
chapters (17)
- 0:00Introduction and context
- 2:17The case for building
- 4:20Hidden costs of building
- 7:24Sustainability and outages
- 9:05The Dilophosaurus factor
- 11:46Calculating development costs
- 14:05Focusing on core competency
- 15:38Organizational maturity
- 18:05The reality of buying
- 20:53Implementation challenges
- 23:35Vendor and process integration
- 26:46Vendor lock-in
- 29:35Not invented here syndrome
- 32:48Five questions for evaluation
- 36:03Bias and sustainability
- 39:06Applying lessons for the future
- 46:23Closing remarks
related talks
35mWe Didn’t Stop to Ask If We Should: Understanding Build vs. Buy
28mPostmortem on the Ingen Incident
44mPitfalls in Technology Selection
30mBuild vs. Buy - a strategic decision for systems and people
11mMaking better build vs buy decisions
48mBuild vs. Buy: The enterprise software conundrum
46mBuild, buy, or use open source? All your answers in one chart
30mHow we chose between building vs buying software