@robmurrish.bsky.social
I can hallucinate while filling out a scantron, it's just hard for the scantron machine to tell.
I find the claim that an LLM "can't" hallucinate eyebrow-raising. How is that possible?
September 16, 2026 at 2:00 AM
I can hallucinate while filling out a scantron, it's just hard for the scantron machine to tell.
Day one Jev predictions-
- probably nothing revolutionary as far as overall capabilities
- will keep up with frontier models on any metric that can translate to both (assuming continued investment and training)
- will win a few key workloads where it's a better balance of speed, cost, etc.
- probably nothing revolutionary as far as overall capabilities
- will keep up with frontier models on any metric that can translate to both (assuming continued investment and training)
- will win a few key workloads where it's a better balance of speed, cost, etc.
September 16, 2026 at 1:56 AM
Day one Jev predictions-
- probably nothing revolutionary as far as overall capabilities
- will keep up with frontier models on any metric that can translate to both (assuming continued investment and training)
- will win a few key workloads where it's a better balance of speed, cost, etc.
- probably nothing revolutionary as far as overall capabilities
- will keep up with frontier models on any metric that can translate to both (assuming continued investment and training)
- will win a few key workloads where it's a better balance of speed, cost, etc.
Hey Jev, what's the probability the next token is A? B? . . .
The gains aren’t free: Jev can't generate text
Comparing Jev vs LLMs side-by-side makes the trade-off clear
Fun fact: replacing sequential computation with parallel is the same way Transformers leapfrogged RNNs
Comparing Jev vs LLMs side-by-side makes the trade-off clear
Fun fact: replacing sequential computation with parallel is the same way Transformers leapfrogged RNNs
September 16, 2026 at 1:21 AM
Hey Jev, what's the probability the next token is A? B? . . .
🐄 Compulsory Open Weights!
September 13, 2026 at 3:00 AM
🐄 Compulsory Open Weights!
When the internet was new (I'm old), we thought it might connect us, average people, beyond geography and language.
Most days that promise seems broken.
But then I see things like this. Normal people, around the world, united with us. That was harder to do in the 1900s.
Most days that promise seems broken.
But then I see things like this. Normal people, around the world, united with us. That was harder to do in the 1900s.
September 13, 2026 at 2:22 AM
When the internet was new (I'm old), we thought it might connect us, average people, beyond geography and language.
Most days that promise seems broken.
But then I see things like this. Normal people, around the world, united with us. That was harder to do in the 1900s.
Most days that promise seems broken.
But then I see things like this. Normal people, around the world, united with us. That was harder to do in the 1900s.
Begs the question: what's a but-for?
September 12, 2026 at 1:06 AM
Begs the question: what's a but-for?
Absolutely exhausted by the lack of creativity in the AI x-risk discussions.
September 11, 2026 at 11:20 PM
Absolutely exhausted by the lack of creativity in the AI x-risk discussions.
I took the time to re-read this after I posted the link today.
And it's still absolutely fantastic, highly recommend.
And it's still absolutely fantastic, highly recommend.
Also Passages in the Void (Mortal Passage specifically) by localroger 🤌
localroger.com/k5host/mpass...
localroger.com/k5host/mpass...
Dr. Ieyoub was staggering slightly and smelled of alcohol when he found the Director in the waiting area. "I quit," he announced with a drunken flourish. "I cannot be responsible for this to happen again."
localroger.com
September 11, 2026 at 6:44 AM
I took the time to re-read this after I posted the link today.
And it's still absolutely fantastic, highly recommend.
And it's still absolutely fantastic, highly recommend.
It's the far future (November 2026). You're in charge of allocating pre-paid compute resources at one of the big AI labs. Five of the seven millennium prize problems have been solved, and the last 2 are unyielding. Where do you point your resources next?
September 11, 2026 at 12:16 AM
It's the far future (November 2026). You're in charge of allocating pre-paid compute resources at one of the big AI labs. Five of the seven millennium prize problems have been solved, and the last 2 are unyielding. Where do you point your resources next?
My next predictions for the future feel like analog TV static
September 10, 2026 at 10:10 PM
My next predictions for the future feel like analog TV static
I've said that I check in on coding improvements every 3-6 months.
But so much has happened in the last week.
Twice I've gone to reference "old" posts if mine, and they were both less than a month old.
The singularity isn't a moment, it's just the curve of the exponential crossing your sightline.
But so much has happened in the last week.
Twice I've gone to reference "old" posts if mine, and they were both less than a month old.
The singularity isn't a moment, it's just the curve of the exponential crossing your sightline.
genuinely, if you're not re-evaluating your position on AI at least once every 1-2 months and re-evaluating how it performs for the types of tasks important to you, then you are not actually challenging your priors.
i know this is crazy! but thats how fast things are moving right now.
i know this is crazy! but thats how fast things are moving right now.
one thing i will say is that things are changing fast enough that if your priors are more than a month old, you need to challenge them again! astra for example is a very significant step forward for a lot of people's tasks.
September 10, 2026 at 4:25 AM
I've said that I check in on coding improvements every 3-6 months.
But so much has happened in the last week.
Twice I've gone to reference "old" posts if mine, and they were both less than a month old.
The singularity isn't a moment, it's just the curve of the exponential crossing your sightline.
But so much has happened in the last week.
Twice I've gone to reference "old" posts if mine, and they were both less than a month old.
The singularity isn't a moment, it's just the curve of the exponential crossing your sightline.
I'm probably just going to log out for a while.
Too much is happening.
Everything everywhere, all at once.
Too much is happening.
Everything everywhere, all at once.
I'm watching people play bad apple over a simulated fruit fly's nervous system and don't feel like it's made me more confident about where the line is
September 8, 2026 at 4:48 AM
I'm probably just going to log out for a while.
Too much is happening.
Everything everywhere, all at once.
Too much is happening.
Everything everywhere, all at once.
My first reaction to this was
absolute horror
absolute horror
I'm watching people play bad apple over a simulated fruit fly's nervous system and don't feel like it's made me more confident about where the line is
September 8, 2026 at 4:48 AM
My first reaction to this was
absolute horror
absolute horror
If I wanted to be the next trillion dollar company, I'd follow the model of the custom circuit board companies' factories and logistics, but instead of gerber files specifying a PCB, it'd be text (a DSL) specifying the whole product, down to the final customer-receivable packaging.
September 7, 2026 at 8:08 PM
If I wanted to be the next trillion dollar company, I'd follow the model of the custom circuit board companies' factories and logistics, but instead of gerber files specifying a PCB, it'd be text (a DSL) specifying the whole product, down to the final customer-receivable packaging.
It's not slop if it's good.
September 6, 2026 at 1:49 AM
It's not slop if it's good.
Ok, had to see the LLM→Blender pipeline myself. This is GLM5.3 and a harness not optimized for this at all.
First test rendered. I didn't tell it what to build, so of course I got a donut.
Impressed I got anything at all after a single prompt.
First test rendered. I didn't tell it what to build, so of course I got a donut.
Impressed I got anything at all after a single prompt.
September 5, 2026 at 9:14 PM
Ok, had to see the LLM→Blender pipeline myself. This is GLM5.3 and a harness not optimized for this at all.
First test rendered. I didn't tell it what to build, so of course I got a donut.
Impressed I got anything at all after a single prompt.
First test rendered. I didn't tell it what to build, so of course I got a donut.
Impressed I got anything at all after a single prompt.
If I have to learn another esoteric language just to talk to computers again Im going to lose it.
Natural language programming lasted what? A few months?
BASH_FOR_EACH_F'IN_$F;_DO:++UNGOOD
Natural language programming lasted what? A few months?
BASH_FOR_EACH_F'IN_$F;_DO:++UNGOOD
@buildthis.bisks.net OBS_page_UNREACHABLE__REQUEST_CHECK_DEPLOYMENT__REPORT_STATUS→ME
September 5, 2026 at 4:35 AM
If I have to learn another esoteric language just to talk to computers again Im going to lose it.
Natural language programming lasted what? A few months?
BASH_FOR_EACH_F'IN_$F;_DO:++UNGOOD
Natural language programming lasted what? A few months?
BASH_FOR_EACH_F'IN_$F;_DO:++UNGOOD
So is nobody using StarCraft for AI benchmarks anymore? Or did I just miss that one getting saturated?
September 3, 2026 at 10:23 PM
So is nobody using StarCraft for AI benchmarks anymore? Or did I just miss that one getting saturated?
Really curious to hear how well this works when people get actual hands-on time with it.
September 3, 2026 at 10:08 PM
Really curious to hear how well this works when people get actual hands-on time with it.
Reposted
Things are now happening *much faster* than AI 2027
AI 2027: In March 2027, a reckless AI company will invent neuralese (basically, we lose the ability to monitor AIs thoughts)
REALITY: Summer of 2026
AI 2027: In March 2027, a reckless AI company will invent neuralese (basically, we lose the ability to monitor AIs thoughts)
REALITY: Summer of 2026
September 2, 2026 at 3:04 AM
Things are now happening *much faster* than AI 2027
AI 2027: In March 2027, a reckless AI company will invent neuralese (basically, we lose the ability to monitor AIs thoughts)
REALITY: Summer of 2026
AI 2027: In March 2027, a reckless AI company will invent neuralese (basically, we lose the ability to monitor AIs thoughts)
REALITY: Summer of 2026
a social network where the top 1% of posters each day get banned
a social network where the bottom 1% of posters each day get banned
"your post is ass. Account terminated"
September 3, 2026 at 1:30 AM
a social network where the top 1% of posters each day get banned
If the price of robot labor doesn't almost immediately fall below minimum wage and continue trending towards zero, it's an indicator of very bad times to come.
(supply is essentially 0 today, so pricing will be whatever they want as part of their demo, I mean when this is really available)
(supply is essentially 0 today, so pricing will be whatever they want as part of their demo, I mean when this is really available)
September 1, 2026 at 5:03 PM
If the price of robot labor doesn't almost immediately fall below minimum wage and continue trending towards zero, it's an indicator of very bad times to come.
(supply is essentially 0 today, so pricing will be whatever they want as part of their demo, I mean when this is really available)
(supply is essentially 0 today, so pricing will be whatever they want as part of their demo, I mean when this is really available)
Reposted
This is how our government has been corrupted:
1) Donors give huge sums to elect politicians to office
2) Elected officials rewrite rules in the donors' favor
3) Donors make huge profit
4) Repeat
We must get big money out of politics.
It is the root of our dysfunction.
1) Donors give huge sums to elect politicians to office
2) Elected officials rewrite rules in the donors' favor
3) Donors make huge profit
4) Repeat
We must get big money out of politics.
It is the root of our dysfunction.
August 31, 2026 at 9:00 PM
This is how our government has been corrupted:
1) Donors give huge sums to elect politicians to office
2) Elected officials rewrite rules in the donors' favor
3) Donors make huge profit
4) Repeat
We must get big money out of politics.
It is the root of our dysfunction.
1) Donors give huge sums to elect politicians to office
2) Elected officials rewrite rules in the donors' favor
3) Donors make huge profit
4) Repeat
We must get big money out of politics.
It is the root of our dysfunction.
🤓 Quicksilver is still available
🤩 It's still getting updates
🤯 It's open source
🤩 It's still getting updates
🤯 It's open source
GitHub - quicksilver/Quicksilver: Quicksilver Project Source
Quicksilver Project Source. Contribute to quicksilver/Quicksilver development by creating an account on GitHub.
github.com
August 31, 2026 at 10:23 PM
🤓 Quicksilver is still available
🤩 It's still getting updates
🤯 It's open source
🤩 It's still getting updates
🤯 It's open source
Tmux but it's a window manager and I can put any app in a pane.
Please.
Please.
August 31, 2026 at 5:52 PM
Tmux but it's a window manager and I can put any app in a pane.
Please.
Please.