Published inTowards Data ScienceTemperature Scaling and Beam Search Text Generation in LLMs, for the ML-AdjacentWhat “temperature” is, how it works, its relationship to the beam search heuristic, and where LLM output generation can still failApr 263Apr 263
The AREA Model of OwnershipAccountability, Responsibility, Expertise and Authorization for code, data and other stuffMar 281Mar 281
Published inTowards Data SciencePyTorch and MLX for Apple SiliconA side-by-side CNN implementation and comparisonMar 81Mar 81
Examining C2PA Provenance Metadata in DALL·E 3 ImagesInspecting the new metadata fields added to synthetically-generated images from ChatGPTFeb 17Feb 17
Platform Monopolies and the Decay of LeverageWhat happens to poorly-run software platforms over time?Feb 11Feb 11
The Simple Elegance of Priority Stack RankingTeams are already prioritizing their roadmaps; but often they’re doing it without consistent guidance and with x-communication overheadJan 29Jan 29
Strategy, Alignment, and Organizational EntropyA given “tops-down” or “bottoms-up” strategy is already a failureDec 15, 20231Dec 15, 20231