Posts

TIL how to setup my terminal to open a file in VS Code with `code`.

On some struggles with chatbots and gensearch tools.

Learned how to use git status --short and create git aliases to filter out _site directory changes from Jekyll builds. Includes search tool comparisons using You.com, Perplexity AI, and Phind.

"A nanopublication is the smallest unit of publishable information."

An example setup for transcribing videos from TikTok.

A thread from Google's Search Liaison.

Screenshot of Ayhan Fuat Çelik's "The Fall of Stack Overflow" on Observable omitted. The graph in question has since been updated. \@natfriedman via Twitter on Jul 26, 2023 Why the precipitous sudden...

This is partially about prompt engineering and partially about what a good essay or search does. More than answer a question, perhaps? (this is engaged with in the essay, though not to my liking). Gri...

\@aravsrinivas via Twitter on Jul 24, 2023 The ultimate question is what is the question. Asking the right question is hard. Even framing a question is hard. Hence why at perplexity, we don’t just let...

A provocation.

Documentation of the LangChain Hub as a searchable repository of prompts for LLM developers, with investigation into when search functionality was added.

The jagged frontier of AI capability isn't random. It's predictable from the practices, artifacts, and feedback loops that constitute actual work. The frontier is smooth where work is extended. It's jagged where work has been stripped to an isolated task.

Each extension has specific capacities. The compiler grounds 'does it run?' A security review grounds 'is it safe?' Map the capacities of the extensions in a practice, and you've mapped where the frontier is smooth and where it isn't. The handoff analytic reveals what happens when those capacities shift.

Productivity and quality are one outcome among many. When AI enters a practice, skill formation, craft, pace, accountability, and repairability all shift. The handoff analytic is what makes these dimensions visible.

The AI discourse obsesses over output quality. But for most of life, the harder problem is upstream—how do you turn 'something feels wrong' into a question worth asking?

Task-exposure models count what AI can do. Bundle theory asks whether the task can be separated from the job. The extensions framework asks a different question: does 'can do' mean the same thing with and without the practice's feedback loops?

Verification catches the error. Repairability determines whether you can do anything about it. They're different properties of the work, and repairability is itself an extension—one that can be designed in or absent.

Cross-trained practitioners—doctors who code, lawyers who build tools—see the frontier differently because they carry extensions from one domain into another. Their accounts reveal the mechanism: it's not the model that differs, it's what the practice provides.

If the extensions framework is right, it should be falsifiable. Here's the specific within-domain prediction: the same task, in the same domain, will show different AI performance depending on whether the practice's extensions are engaged. Not across domains—within one.

The hard problems remain, how to use a tool to do something you or someone else wants.