AI

Tesla finally releases FSD v12, its last hope for self-driving | Electrek - 300k rules based system down to a model - you’ll need both: a model and overrides. - Or, you train the model by learning from the scenarios encoded in your 300k rules engine (like DeepMind’s AlphaGo)

Fast Tracking RAG Model Evaluation with Synthetic Data - synthetic data to evaluate RAG responses? - A response is non-deterministic, random - Perhaps do a Levenshtein distance of sorts to evaluate of response is as good as you want? - But isn’t this the same as question-answer pairs and instruction tuning?

Auto Wiki by Mutable.ai - this proves how powerful AI is. Just add data and watch it churn. - An example of an AI product.

Coding

Python Does Not Fully Deliver OOP Encapsulation Functionalities - I learned how attributes are names in a Python class. As an R programmer, most of these things were never taught to me in school, and I’ve never really needed to know them (until working with software engineers)

Product

How Gong builds product - by Lenny Rachitsky - the planning!! > We do not plan monthly or biweekly. Both the engineering leader and I dislike the Scrum methodology. We feel it’s trying to drive urgency via artificial deadlines versus via value to the customer. And by forcing “commitment” to deliverables within a time window, it essentially inhibits on-the-fly trade-offs between content, quality, and timelines. Yet we have internal reviews with the different groups on a monthly basis.

_________________________

Bryan lives somewhere at the intersection of faith, fatherhood, and futurism and writes about tech, books, Christianity, gratitude, and whatever’s on his mind. If you liked reading, perhaps you’ll also like subscribing: