An LLM wiki can turn saved research into a linked, searchable collection that grows as you add sources. Journalist Casey Newton says his version helped him recover context on a fast-moving story and spot a possible podcast topic—but it also required frequent maintenance, and he stopped short of recommending the setup broadly. His experience is a useful case study, not proof that an LLM wiki makes everyone more productive.
What Newton built
In an August 19, 2026, account for Platformer, Newton describes creating a research wiki from Markdown files in Obsidian. The idea followed Andrej Karpathy’s observation, quoted in Newton’s essay: “Something I’m finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest.”
As an Amazon Associate I earn from qualifying purchases.
Newton used Claude Code to create a prompt for a “Karpathy-style LLM wiki,” then ran it in a terminal. He initially populated the wiki with his Platformer writing archive dating back to 2020. For ongoing additions, he saves stories with Obsidian’s web clipper, which converts them to Markdown; a local script reads the new material and determines where it belongs. An Obsidian community plugin lets him ask questions across the collection.
The system builds detailed timelines linked to original sources, generates pages for concepts appearing in the news, and refreshes a home page each morning. Newton reported having more than 1,440 wiki pages when he wrote the essay. His Meta page was more than 12,000 words long and linked to over 1,300 other wiki pages. Those figures describe his own 2026 setup; they are not independently audited measures of its usefulness.
#1 Best Overall
What it helped him do
Recover context on a developing story
When reporting on a fast-moving story about an OpenAI and Hugging Face agentic breach, Newton used the wiki to recall developments. He says he then opened the original sources to check that he was not about to repeat a hallucination. The distinction matters: the wiki helped him find and reassemble context, but source verification remained part of the work.
Notice a possible topic
Newly generated concept pages also helped him identify a possible podcast topic about AI and Congress. That is an example of idea discovery in his workflow, not evidence that an LLM wiki will reliably generate useful story ideas for other users.
Replace workflows that were hard to sustain
Newton had previously tried a Notion agent-search feature that took effort to use and sometimes failed to cite sources without extra prompting. He also maintained “blips” in Capacities, but keeping those notes current became difficult as they multiplied. His wiki offered a different balance: more automated linking and updating, in exchange for ongoing technical and editorial upkeep.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
How to decide whether the approach fits
The practical question is not simply whether an LLM can summarize notes. It is whether the whole pipeline—from saving source material to checking generated pages—fits the way you work. Consider these factors before building one:
- Archive size and value: A large body of material that you repeatedly revisit may justify indexing more than a small collection you rarely search.
- Source traceability: Check whether generated claims remain connected to the original article, document, or other source, and whether you can open that source easily to verify details.
- Automation needs: Newton’s setup uses a clipper, a local script, and a community plugin. Each adds capability, but also another part of the workflow that may need configuration or repair.
- Maintenance tolerance: The system is a poor fit if you expect a knowledge base to stay accurate and orderly without regular review.
- Compatibility with existing habits: A new wiki is only useful if saving, finding, and reviewing material feels easier than maintaining the notes or search system you already use.
The upkeep Newton encountered
Pages need active editing
Newton says the wiki needs “more or less constant maintenance.” Pages can become too long and need to be compacted. Generated text also required editing: he found Claude’s compressed style unsuitable for his needs and used another model to rewrite much of the output closer to AP style.
Automation can stop working
Code errors sometimes halted a process. That means the system’s convenience depends on a working chain of tools and scripts, and someone must notice when an update fails rather than assume the wiki has kept itself current.
Rank #4
There is no general productivity verdict
Newton called the wiki the most useful productivity change he tried that year, while also describing it as too specific and clunky to confidently recommend to others. His account does not establish a measured productivity gain or show that the benefits outweigh the maintenance for users with different archives, tools, or research routines.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsWhat to take from the case study
Newton’s workflow shows one way to turn saved sources into linked Markdown pages that can be updated and queried. Its clearest value in his account was faster retrieval of prior context and the chance to notice a connection between topics. Its costs were ongoing page cleanup, troubleshooting, and editing generated prose. Treat the setup as a possible fit for research-heavy work—not a plug-and-play productivity upgrade—and keep original sources close enough to verify what the wiki tells you.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
