nickspag.bsky.social
@nickspag.bsky.social
Mmm post-prep?
November 27, 2025 at 8:10 PM
Mmm post-prep?
Kinda hard to tell. Compute and DC spending still indicates speculative confidence. There’s signs we’re nearing the “horsepower” limits of the current LLM architecture. but we’ve barely even begun to integrate them in to existing processes, so still lots of potential for labor disruption, etc.
November 19, 2025 at 9:28 PM
Kinda hard to tell. Compute and DC spending still indicates speculative confidence. There’s signs we’re nearing the “horsepower” limits of the current LLM architecture. but we’ve barely even begun to integrate them in to existing processes, so still lots of potential for labor disruption, etc.
All data is synthetic. And humans themselves are trained on tons of it.
November 7, 2025 at 6:30 PM
All data is synthetic. And humans themselves are trained on tons of it.
You’re also trained on synthetic data though! That’s probably the issue
November 7, 2025 at 6:28 PM
You’re also trained on synthetic data though! That’s probably the issue
I think it’d be the opposite these days
August 8, 2025 at 5:55 PM
I think it’d be the opposite these days
Actually this has a bit more on what they actually saw across model families etc arxiv.org/abs/2506.11613
Model Organisms for Emergent Misalignment
Recent work discovered Emergent Misalignment (EM): fine-tuning large language models on narrowly harmful datasets can lead them to become broadly misaligned. A survey of experts prior to publication r...
arxiv.org
July 1, 2025 at 9:01 PM
Actually this has a bit more on what they actually saw across model families etc arxiv.org/abs/2506.11613
Mandating businesses and payroll service providers not withhold? I’m really curious about the potential avenues here as well.
June 7, 2025 at 2:14 PM
Mandating businesses and payroll service providers not withhold? I’m really curious about the potential avenues here as well.
Please point me to any research, would love to be proven wrong.
May 23, 2025 at 2:25 AM
Please point me to any research, would love to be proven wrong.
It’s funny but these detectors are not capable in any way.
May 23, 2025 at 1:21 AM
It’s funny but these detectors are not capable in any way.
I’m being flippant but there’s some really great discussion here with the then-comptroller of the Navy on audits podcasts.apple.com/us/podcast/w...
podcasts.apple.com
April 23, 2025 at 7:09 PM
I’m being flippant but there’s some really great discussion here with the then-comptroller of the Navy on audits podcasts.apple.com/us/podcast/w...
Yea and good luck determining the real value of a carrier strike group.
April 23, 2025 at 6:59 PM
Yea and good luck determining the real value of a carrier strike group.
Yea agreed. And most of the top U.S. cabinet positions actually map quite well to the top broad concerns of voters. A U.S. version could also be more of a party messaging and organization role, and help build out the policy and candidate bench.
February 13, 2025 at 6:16 PM
Yea agreed. And most of the top U.S. cabinet positions actually map quite well to the top broad concerns of voters. A U.S. version could also be more of a party messaging and organization role, and help build out the policy and candidate bench.
Basically, yes. en.m.wikipedia.org/wiki/Shadow_...
Shadow cabinet - Wikipedia
en.m.wikipedia.org
February 13, 2025 at 5:57 PM
Basically, yes. en.m.wikipedia.org/wiki/Shadow_...
Disappointing feels strong but he is really starting to flex the muscle this year and it’s impressive
January 27, 2025 at 12:08 AM
Disappointing feels strong but he is really starting to flex the muscle this year and it’s impressive
God that does not sound sustainable. This may be naive but can you nap at all? Or is it just 12 hours on?
December 10, 2024 at 1:08 AM
God that does not sound sustainable. This may be naive but can you nap at all? Or is it just 12 hours on?
Is there truth to the “most mistakes happen because shift changes so we make the shifts long” rationale?
December 7, 2024 at 8:16 PM
Is there truth to the “most mistakes happen because shift changes so we make the shifts long” rationale?
Is it possible that we just haven’t seen this quick and coupled price/wage increase environment before?
November 12, 2024 at 2:33 PM
Is it possible that we just haven’t seen this quick and coupled price/wage increase environment before?