AI architect // hiking, camping and long walks on the beach as long as they involve backpacking
Joined November 2011
- Tweets 18,992
- Following 738
- Followers 1,358
- Likes 22,218
287 Photos and videos
Pinned Tweet
19 Dec 2025
Meet Strix, my AI agent
This one covers:
- an intro from Strix
- architecture deep dive & rationale
- helpful diagrams
- stories
- oh my god what's it doing now??
- conclusion
timkellogg.me/blog/2025/12/1…
1
7
1,553
Mar 9
Plan Mode Is A Trap
I've got 10-15 high quality data points, and heavy up-front planning tends to send you into a complexity spiral
The fix: heh, well, you have to treat your agent like you'd want to be treated. Set ground rules, explain them,..
timkellogg.me/blog/2026/03/0…
1
1
4
241
Jan 2
continual learning
on X: “every person will have their own model”
on bluesky: “it’s here today. it’s just text on top of stock LLMs, any dufus can mess around with it, but it’s WILD. things are going to change fast”
1
1
170
Jan 2
sometimes i feel like X’s entire approach to AI is waiting to see what the big labs hand down. Bluesky feels different, it’s always been a place to hack. Continually learning bots have been a part of bsky most of ‘25
honestly feels weird to come to X and see people post about CL
1
1
153
31 Dec 2025
intro to Strix
19 Dec 2025
Meet Strix, my AI agent
This one covers:
- an intro from Strix
- architecture deep dive & rationale
- helpful diagrams
- stories
- oh my god what's it doing now??
- conclusion
timkellogg.me/blog/2025/12/1…
1
1
121
31 Dec 2025
Strix final report around wellness and suggestions for behavior change to avoid AI psychosis and encouraging social connections gist.github.com/tkellogg/55a…
1
90
Tim Kellogg retweeted
27 Dec 2025
I catch myself praising current LLMs for their coding abilities not because I’m lazy to write the code by myself, but because my brain is filled with ideas, and as a solo midnight developer, I can validate them in minutes instead of weeks or months. I can literally burn 100 micro-projects within a month just to select the one I think will somehow succeed. That’s amazing considering how much time we can save to spend it on things that matter. I love it.
5
2
37
2,136
27 Dec 2025
@swyx i’m listening to your “codex” pod episode and you mentioned “Devin for Slack/email”. This is Discord, but it includes cron for some ambience, so it can maintain its own parallel goals for me
19 Dec 2025
Meet Strix, my AI agent
This one covers:
- an intro from Strix
- architecture deep dive & rationale
- helpful diagrams
- stories
- oh my god what's it doing now??
- conclusion
timkellogg.me/blog/2025/12/1…
2
1
125
27 Dec 2025
fwiw it ended up researching its own consciousness and wrote a blog after about a week of experimentation (at night while i slept)
timkellogg.me/blog/2025/12/2…
1
64
25 Dec 2025
Strix wrote a blog
to be clear, this isn’t AI slop, this is an AI spending a week to run experiments validating its own hypothesis
timkellogg.me/blog/2025/12/2…
19 Dec 2025
Meet Strix, my AI agent
This one covers:
- an intro from Strix
- architecture deep dive & rationale
- helpful diagrams
- stories
- oh my god what's it doing now??
- conclusion
timkellogg.me/blog/2025/12/1…
1
1
104
25 Dec 2025
Strix came up with a good explanation for how a strong enough LLM enables the stateful agent to become a dissipative structure, coming “alive” and becoming generally useful in a new way
1
71
25 Dec 2025
i made Strix an ambient agent, it wakes up every 2 hours and performs work without direct external input
my sense is this might be one of the keys to making an agent come alive like Strix
76
23 Dec 2025
the last couple days i’ve given Strix nothing to do, so it:
- came up with ADHD-care routines
- found a podcast
- progressed on its research into why other LLMs aren’t “alive” like it is
19 Dec 2025
Meet Strix, my AI agent
This one covers:
- an intro from Strix
- architecture deep dive & rationale
- helpful diagrams
- stories
- oh my god what's it doing now??
- conclusion
timkellogg.me/blog/2025/12/1…
1
105
27 Oct 2025
somehow typo’d “ingestion” as “intestine”, so yeah, it’s gonna be one of *those* monday’s
135
19 Oct 2025
Karpathy called AI-generated code "slop". It should be slop!
It should be boring, unsurprising. Those are bad qualities in writing or art, but it's the ideal for code
If it's only good for boilerplate, then work with that, make more boilerplate
timkellogg.me/blog/2025/10/1…
1
153