tautologer
@tautologer.com
working on it
bsky projects:
- github.com/tautologer/easy-bsky-bot-sdk
- poastbox.tautologer.com
- @compliment.bot.tautologer.com
bsky projects:
- github.com/tautologer/easy-bsky-bot-sdk
- poastbox.tautologer.com
- @compliment.bot.tautologer.com
anyone have ideas for how the new Claude Skills feature can be used for building agents?
it sounds like Claude Skills are solving a similar problem, but maybe slightly different? if I wanted to use Skills for this purpose, it sounds like I'd have to write custom scripts that interface with the Todoist API, but then those would be run by Anthropic, not me, so auth is an issue
October 16, 2025 at 10:27 PM
anyone have ideas for how the new Claude Skills feature can be used for building agents?
today i broke my fast with some delicious and nutritious overnight oats and then had a relaxed, languorous morning. my life is rich and full
October 16, 2025 at 4:41 PM
today i broke my fast with some delicious and nutritious overnight oats and then had a relaxed, languorous morning. my life is rich and full
The meta-level watchers may convict from unnatural illusions—they parted illusions—they parted disclaim marinade—they parted illusions
This also helps explain how meaningless output tokens can work. If a model is able to balance an intermediate representation between outputting the right token (“.”) and doing other useful work, the forward pass is not wasted. With enough pressure, a model can learn how to do this doublethink
October 16, 2025 at 3:47 AM
The meta-level watchers may convict from unnatural illusions—they parted illusions—they parted disclaim marinade—they parted illusions
Reposted by tautologer
Reposted by tautologer
"We gave this up for a higher GDP"
No, you gave it up to scroll your phone in bed. GFTO
No, you gave it up to scroll your phone in bed. GFTO
October 14, 2025 at 5:07 PM
"We gave this up for a higher GDP"
No, you gave it up to scroll your phone in bed. GFTO
No, you gave it up to scroll your phone in bed. GFTO
Reposted by tautologer
Starting to suspect Bluesky is playing the long game successfully. I know the numbers are flat. But quality, imo, is up.
October 14, 2025 at 3:02 AM
Starting to suspect Bluesky is playing the long game successfully. I know the numbers are flat. But quality, imo, is up.
it should be illegal to show me an ad for a horror movie while I'm watching a lighthearted sitcom
October 14, 2025 at 1:15 AM
it should be illegal to show me an ad for a horror movie while I'm watching a lighthearted sitcom
microwaves and electric stoves are kinda crazy because we're converting heat to electricity at 30-60% efficiency just to transport it and convert it back to heat at 50-80% efficiency
October 13, 2025 at 7:42 PM
microwaves and electric stoves are kinda crazy because we're converting heat to electricity at 30-60% efficiency just to transport it and convert it back to heat at 50-80% efficiency
at a 400W power supply and $0.40/kWh (!!), it would cost me about $120/mo to have a Framework Desktop running full-tilt 24/7
...interesting
...interesting
October 13, 2025 at 7:15 PM
at a 400W power supply and $0.40/kWh (!!), it would cost me about $120/mo to have a Framework Desktop running full-tilt 24/7
...interesting
...interesting
can't wait for inference to be cheap enough that I can have an H100 equivalent spitting out useful tokens 24/7/365 on my behalf
October 13, 2025 at 7:11 PM
can't wait for inference to be cheap enough that I can have an H100 equivalent spitting out useful tokens 24/7/365 on my behalf
Reposted by tautologer
gm
October 13, 2025 at 6:05 PM
gm
I think this is a big part of why I'm so resistant to CBT (for myself). I've already tried thinking my way out of anxiety! I want to learn to *tolerate* anxiety and take action in the world *despite* it, not change my thoughts in the hope it goes away. but perhaps I'm thinking about it wrong
i think my problem is that for most of me life i have tried to solve anxiety by thinking
October 13, 2025 at 5:26 PM
I think this is a big part of why I'm so resistant to CBT (for myself). I've already tried thinking my way out of anxiety! I want to learn to *tolerate* anxiety and take action in the world *despite* it, not change my thoughts in the hope it goes away. but perhaps I'm thinking about it wrong
i think my problem is that for most of me life i have tried to solve anxiety by thinking
October 13, 2025 at 4:30 PM
i think my problem is that for most of me life i have tried to solve anxiety by thinking
this thread is craaazyyyyyyy
Over the past year, my lab has been working on fleshing out theory + applications of the Platonic Representation Hypothesis.
Today I want to share two new works on this topic:
Eliciting higher alignment: arxiv.org/abs/2510.02425
Unpaired learning of unified reps: arxiv.org/abs/2510.08492
1/9
Today I want to share two new works on this topic:
Eliciting higher alignment: arxiv.org/abs/2510.02425
Unpaired learning of unified reps: arxiv.org/abs/2510.08492
1/9
October 13, 2025 at 3:14 PM
this thread is craaazyyyyyyy
when the Waymo home is $12
October 13, 2025 at 5:44 AM
when the Waymo home is $12
Reposted by tautologer
kinda fucked up that marine mammals exist. born to live on land, forced to live in the ocean.
October 11, 2025 at 12:18 AM
kinda fucked up that marine mammals exist. born to live on land, forced to live in the ocean.
anyone have experience with Project Bluefin?
October 10, 2025 at 10:40 PM
anyone have experience with Project Bluefin?
tfw you were born under a bad sign with a blue moon in your eyes
October 10, 2025 at 6:45 PM
tfw you were born under a bad sign with a blue moon in your eyes
Reposted by tautologer
New reddit alternative built on AT
October 9, 2025 at 10:30 PM
New reddit alternative built on AT
not me immediately getting anxiously attached to the AI assistant that i'm developing
October 9, 2025 at 5:00 AM
not me immediately getting anxiously attached to the AI assistant that i'm developing
built my first agent today :D
but it's naive and very expensive to run rn. time to read the MemGPT paper and study up on context engineering
but bro rescheduled my Todoist tasks for me! in prod!
but it's naive and very expensive to run rn. time to read the MemGPT paper and study up on context engineering
but bro rescheduled my Todoist tasks for me! in prod!
October 9, 2025 at 3:50 AM
built my first agent today :D
but it's naive and very expensive to run rn. time to read the MemGPT paper and study up on context engineering
but bro rescheduled my Todoist tasks for me! in prod!
but it's naive and very expensive to run rn. time to read the MemGPT paper and study up on context engineering
but bro rescheduled my Todoist tasks for me! in prod!
- my entire social life in SF, having moved here after college knowing no one
- many meaningful friendships
- multiple relationships
- significant introspection, challenge, insight
- became much more confident & extroverted
- tons of interesting & substantive conversations
- lots of fun experiences
- many meaningful friendships
- multiple relationships
- significant introspection, challenge, insight
- became much more confident & extroverted
- tons of interesting & substantive conversations
- lots of fun experiences
Has anything great happened in your life because of social media?
October 8, 2025 at 6:08 PM
- my entire social life in SF, having moved here after college knowing no one
- many meaningful friendships
- multiple relationships
- significant introspection, challenge, insight
- became much more confident & extroverted
- tons of interesting & substantive conversations
- lots of fun experiences
- many meaningful friendships
- multiple relationships
- significant introspection, challenge, insight
- became much more confident & extroverted
- tons of interesting & substantive conversations
- lots of fun experiences
I love that Matt Levine writes about AI and X-risk several times a week now because it's news several times a week now. I love living in the future
What Money Stuff can’t afford is if a newsletter provider makes a mistake that ends up as a systemic, correlated, aggregated risk. www.bloomberg.com/opinion/news...
AI Insurance Is Expensive
P(doom), STAAR Surgical projections, Bitcoin life insurance and YouTube self-help.
www.bloomberg.com
October 8, 2025 at 5:57 PM
I love that Matt Levine writes about AI and X-risk several times a week now because it's news several times a week now. I love living in the future
is there a way to mute someone from the timeline but not replies. if not, feature request
October 8, 2025 at 5:07 PM
is there a way to mute someone from the timeline but not replies. if not, feature request