---
title: "Tim Kellogg"
type: "entity"
entity_kind: "person"
status: "hub"
canonical_name: "Tim Kellogg"
aliases: []
first_seen: "2026-08-28T00:00:00.000Z"
writer_model: "claude-sonnet-5"
connects_to: ["s1","AI cost estimation","independent AI blogging","source-tier discipline"]
seek_code_commit: "7d6d9ed"
---


Independent AI blogger who publishes technical commentary and analysis on
his own site, timkellogg.me.

Matters to this vault as a case study in the sourcing-tier rubric: two days
before TechCrunch's widely-circulated "under \$50" headline for training
[[claim-s1-distilled-reasoning-from-1000-traces-in-26-minutes|s1]], Kellogg
published his own arithmetic on the same underlying GPU-hours figure from
the paper and arrived at a third, lower estimate — "about \$6" — in a post
titled "S1: The \$6 R1 Competitor?" His figure was earlier, showed its work
more plainly (the title itself hedges with a question mark), and diverged
from the number that ended up traveling — see
[[claim-kellogg-independent-6-dollar-s1-estimate-predates-techcrunch]].

## References
- [[claim-kellogg-independent-6-dollar-s1-estimate-predates-techcrunch]]