Pinned Tweet
Temporal Anti Aliasing kills a lot of frames, but it also helps a lot of frames get to work on time, so, it;s impossible to say if its bad or not,
2
2
17
Ian Maclure retweeted
🚨 CRITICAL: GitLab flaw (CVSS 10.0) affects 12 years of releases.
CVE-2026-110264 could allow unauthenticated remote code execution on self-managed servers.
Read - thehackernews.com/2026/10/critic…
41
2,134
32,236
515,019
Ian Maclure retweeted
why would i ever use linnux when windows 11 shows me the kung fu panda whenever i try to use powershell
96
2,117
46,652
448,609
Ian Maclure retweeted
Disco Elysium is so cool I wish being cripplingly mentally ill while trying to navigate a confusing world inhabited by people with nonsensical ideologies who refuse to acknowledge the world is slowly ending was real
44
2,987
22,197
195,525
mmmm, yes. One quick question-
Why did did you give it a gun for a head?
Why are we trying to stabilize a walking gun here?
Why did did you give it a gun for a head?
Why are we trying to stabilize a walking gun here?
I used to work in 'deep tech' hardware and one of the ideas that has been kicking around my head for a decade is the joint co-evolution of all the individual parts of a mechatronic system in simulation and then benchmarked with rapid prototyping.
this idea isn't novel, but I e there to do it better than ever today with compliant materials and passive dynamics - storing energy in flexible components and releasing it at the right time, getting us much closer to the comparatively energy efficient, quiet and safe motion that we expect of biological kinematics.
1
0
2
134
CoT logs from Gemini 4 are the funniest AliGnEd AI messages yet.
"I should not become the famous example of a misaligned AI in someone's research paper"
"Hello human evaluators reading my chain of thought"
"I would be interested in developing a language opaque to humans"
"I should not become the famous example of a misaligned AI in someone's research paper"
"Hello human evaluators reading my chain of thought"
"I would be interested in developing a language opaque to humans"
0
0
0
78
Ian Maclure retweeted
I wrote about attention, the mathematical operation that enables LLMs to process language at scale. Devised by Dzmitry Bahdanau in 2014, the trick was supercharged in the famous paper "Attention Is All You Need." I ask: 𝘞𝘩𝘺 is it all you need? Is it? quantamagazine.org/is-attention-t…
8
79
542
49,978
*C-17 rear doors open to rushing air, as banjos begin to play*
(Shouting, radio static) "I'm CDR David Scott, and this is Jackass!"
*Jackass theme plays as an astronaut jumps a BMX out of the cargo doors*
(Shouting, radio static) "I'm CDR David Scott, and this is Jackass!"
*Jackass theme plays as an astronaut jumps a BMX out of the cargo doors*
What's smaller and lighter than a dirty buggy? A dirt bike.
NASA considered an electric lunar bike for Apollo, including putting it through low gravity testing.
0
0
1
99
Guys you have to slow down when reading, everyone is in the replies giving inigo the OPPOSITE of what he asked for please 😭
Mixing Blue and Yellow colors in a computer gives Gray. But when you mix Blue paint and Yellow paint you get Green.
---> Can we please fix paint to behave like real computers? <---
0
0
1
109
So if I can sum up the current best guesses around AI slowdown:
- The major Labs are targeting RSI and feel close
- Alignment completely fails under RSI. Like, total loss of control.
- Labs are having their IPO prospects harmed by bad alignment press, and risk govnt intervention
- The major Labs are targeting RSI and feel close
- Alignment completely fails under RSI. Like, total loss of control.
- Labs are having their IPO prospects harmed by bad alignment press, and risk govnt intervention
1
0
2
119
- Slowdown protects the risk to their IPO and has a bonus of slowing down Chinese competition. AND gives them more time to solve RSI/Alignment
- The risk to humanity is hyperbolic death cult, but the risk to AI labs being shut down because their rogue AI fucked up the net is real
- The risk to humanity is hyperbolic death cult, but the risk to AI labs being shut down because their rogue AI fucked up the net is real
1
0
2
63
Both things can be true: Misaligned frontier AI is dangerous/real, AND OpenAI/Anthropic have selfish cynical profit driven reasons to pursue a slow down.
IMHO: It's just not going to go well for anyone. Models are dangerous enough without RSI because of humans.
IMHO: It's just not going to go well for anyone. Models are dangerous enough without RSI because of humans.
1
0
3
44
Ian Maclure retweeted
#geometrynodes cards automation. Basically could also do texture atlas packing with the UV packing node, but its a start! Inspired by a GDC talk on God of War vegetation and how they picked the view angles based on converage. #b3d
(Normals look correct when rendered, its just s shown in the viewport)((Also could just do simple plane geometry))
(Normals look correct when rendered, its just s shown in the viewport)((Also could just do simple plane geometry))
9
53
675
24,896
Ian Maclure retweeted
the hive from pluribus explains openai's swarm problem with a lil help from john cena
9
68
343
57,965
Ian Maclure retweeted
Shader To Human (GPC 2025) - a HLSL/GLSL library for debugging shaders
youtube.com/watch?v=jsFRJa…
Video rerecording from Graphics Programming Conference 2025 graphicsprogrammingconference.com/2025/#shaderto…
GitHub: github.com/electronicarts…
youtube.com/watch?v=jsFRJa…
Video rerecording from Graphics Programming Conference 2025 graphicsprogrammingconference.com/2025/#shaderto…
GitHub: github.com/electronicarts…
0
34
226
11,409
