2 minute read / Jan 1, 0001 /
2024-07-24
- impact on dev tools with AI
- consolidation within developers tools
- new wave of AI dev tools as the others might be commoditized
Service as a SaaS
- toil + labor market discontinuity
Software job market
https://matt.sh/panic-at-the-job-market
Software is a factory.
- Diagram of all the different parts of software and the parts that are being automated much the same way that automobile Manufacturing was productized
Crypto marketing spend
- analysis of the amount of money major projects are using to court developers and drive adoption.
- https://www.coindesk.com/business/2024/07/02/polkadots-245m-treasury-will-last-2-years-at-current-spending-rate/
PayPal stablecoin
- look at the rise of the PayPal stable coin and how it compares to other stable coins.
Model release duration
- look to see if models are taking much longer to release which provides more stability to build between generations
- very much like Google SEO where companies would operate within the confines of one particular model and then when a new model was released it would destabilize the businesses built
- Analogy is SEO where some significant change would change the total number of successful promptss
ideas
-
Changing need in education.
- AI investments by category.
- Value capture of AI as a function of savings.
- Aaron Lovey’s post on AI.
- And Vinod Khosla’s dreams of the future.
- OpenAI versus Google traffic
Still forming
- the book broken money
Books
- the trading book
- education
Trends
Random thoughts
- Quality revenue and software companies.
- sales development, hiring, flattening.
- theory one year in
- financing expecattions
- rise of the series b
- software companies slowing down
- federer speech
- apple making privacy a feature
A rapid commoditization of AI models and their significant capital intensity will mean that training the bigger models is reserved just for the companies who have incredible business models and who can return the invested capital in a relatively short period of time because they have a large customer base that is motivated to move to the newest models and pay a premium. Over time, as the performance asymptotes, it should become less and less attractive to train larger models unless there is a business model to support it.