Tim Kellogg
timkellogg.me · Personal blog · on HN since 2012
Tim Kellogg writes about AI and LLM behavior, agents and protocols like MCP, and general software topics such as regex and HTTP/2.
What HN saysHN commenters cite his experiments on LLM behavior, such as whether models get bored, as evidence that AI shows emergent personality.
29HN people shared it
6linked it in comments
85stories submitted
1.7Kstory points
6 months agolast shared
2012people per year2026
Most-shared posts
- 1Regex Isn't Hard (2023)4 people · 79 points · first shared 2023 · 97 comments on HN
- 2Avoiding Fallback and Cold Paths4 people · 10 points · first shared 2021 · 1 comments on HN
- 3Does AI Get Bored?2 people · 24 points · first shared 2025 · 16 comments on HN
- 4MCP Is Unnecessary2 people · 8 points · first shared 2025 · 1 comments on HN
- 5I Use AI2 people · 6 points · first shared 2025 · 1 comments on HN
- 6Can HTTP/2 Replace MQTT?2 people · 6 points · first shared 2015 · discussion on HN
- 7Recursive Improvement: AI Singularity or Just Benchmark Saturation?2 people · 5 points · first shared 2025 · 1 comments on HN
- 8Why I'm Not Going To Stop Posting Go Links2 people · 5 points · first shared 2014 · 1 comments on HN
- 9What Is Entropix Doing?2 people · 5 points · first shared 2024 · discussion on HN
- 10FP for the Working Programmer: Why Is Null Bad?2 people · 3 points · first shared 2024 · discussion on HN
- 11An Unbiased Comparison of F# and Scala2 people · 3 points · first shared 2013 · discussion on HN
- 12LRMs Are Interpretable2 people · 3 points · first shared 2024 · discussion on HN
- 13The Three Plates Method2 people · 3 points · first shared 2022 · 1 comments on HN
- 14How Emotional Is Trump?2 people · 3 points · first shared 2024 · 1 comments on HN
- 15Target Practice: Resumes, but Better2 people · 2 points · first shared 2025 · discussion on HN
What HN people wrote when they linked it
andy99 · 2026-05-18
I’m definitely not in the “ai is sentient” camp, but it obviously has personality and emergent behaviours including when left to its own devices. There have been various experiments on this e.g. https://timkellogg.me/blog/2025/09/27/boredom
verdverm · 2025-11-25
Simon and Tim have a good thread about this on Bsky: https://bsky.app/profile/timkellogg.me/post/3m4ridhi3ps25
Tim also wrote about this topic: https://timkellogg.me/blog/2025/11/03/colors
Kye · 2025-02-21
A toddler can learn by trial and error mid-process. An LLM using autoregressive inference can only compound errors. The LLDM model paper was posted elsewhere, but: https://arxiv.org/pdf/2502.09992
It basically uses the image generation approach of progressively refining the entire thing at once, but applied to text. It can self-correct mid-process.
The blog post where I found it originally that goes into more detail and raises some issues with it: https://timkellogg.me/blog/2025/02/17/diffusion
gradschoolfail · 2024-09-10
https://timkellogg.me/blog/2024/06/26/politics-analysis
Any luck with the Schuetzenberger thm?
Will get back to aristotle when more bw thanks!
tkellogg · 2024-01-16
I made fossil for this purpose. Sometimes the good stuff is regular social posts, sometimes it’s links. The only way to separate the wheat from the chaff is AI (imo)
https://timkellogg.me/blog/2024/01/03/birb
forgotmypw17 · 2022-04-11