Improving Yesterday
summary
A startup called Grownome builds 3D‑printed IoT sensor nodes on a Raspberry Pi Zero W that stream temperature/humidity and images into a pipeline using Balena, Firebase then Postgres, BigQuery and Google Cloud Functions, training image classifiers in Google AutoML to predict “watered” vs “not watered.” The ML works well, but customer traction is minimal (five paying users), monthly recurring revenue metric stays tiny, cloud credits expire and the project shuts down; specs and Orchestra improve reliability but do not change market fit.
AI-written summary · about
chapters (23)
- 0:00Introduction and startup idea
- 1:06Why build Grownome
- 3:10Building the first prototype
- 4:13Mistakes in business setup
- 5:20Defining value and metrics
- 8:50Customer feedback and kool-aid
- 11:20Startup School and competitors
- 15:39Y Combinator Startup School experience
- 19:28Team dynamics and conflict
- 21:16Rewriting backend architecture
- 23:21Machine learning model for plants
- 24:46Language and tech stack choices
- 26:55Data validation and debugging
- 27:59Database and team communication
- 29:01Building the customer portal
- 30:24Operations and team skills
- 31:48Training and learning gaps
- 32:49Google Stackdriver and monitoring
- 33:33Team communication tools
- 34:17Project management and tools
- 35:19Documentation and knowledge sharing
- 35:59Infrastructure costs and shutdown
- 36:42Key takeaways and summary
related talks
40mRolling out Enterprise AI: Tools, Insights, & Team Empowerment
22mBuild Smarter, Don't Just Build
51mScaling Sustainable Innovation through Team Autonomy
30mSucceeding with Cloud
35mBuilding Data-Intensive Teams
31mManaging Growing Pains: Thinking Big While Being Small
46mData-Driven Transformation: Organizational Changes and Leadership Insights
51mTrade-Offs in Building a Software Company from Startup to Scale Up