@annoyingreposter.bsky.social
Pinned
I might look smart, however, I am absolutely not.
today, I debugged a big chunk of code and hyperparameters (I can't do search thereof) with Gemini (free).
the incentive is here, the model tends to "think" in a right direction, however, me and it more ran in circles than did actual stuff
knowing that, I'm not sure that I want to spend money on AI
the incentive is here, the model tends to "think" in a right direction, however, me and it more ran in circles than did actual stuff
knowing that, I'm not sure that I want to spend money on AI
I still don't feel like I have a good enough answer to the how you get the $2400/yr worth of value back.
I acknowledge that it enables building custom software and generally code faster, that is cool.
The part I can't understand is why people are willing to pay $2400/yr for that ability.
I acknowledge that it enables building custom software and generally code faster, that is cool.
The part I can't understand is why people are willing to pay $2400/yr for that ability.
December 30, 2025 at 4:23 PM
today, I debugged a big chunk of code and hyperparameters (I can't do search thereof) with Gemini (free).
the incentive is here, the model tends to "think" in a right direction, however, me and it more ran in circles than did actual stuff
knowing that, I'm not sure that I want to spend money on AI
the incentive is here, the model tends to "think" in a right direction, however, me and it more ran in circles than did actual stuff
knowing that, I'm not sure that I want to spend money on AI
Reposted
What if you could train agents on a 𝗱𝗲𝗰𝗮𝗱𝗲 of driving experience in 𝘂𝗻𝗱𝗲𝗿 𝗮𝗻 𝗵𝗼𝘂𝗿, on a single GPU?
Excited to share 𝙋𝙪𝙛𝙛𝙚𝙧𝘿𝙧𝙞𝙫𝙚 2.0: A fast, friendly driving simulator with RL training via PufferLib at 𝟯𝟬𝟬𝗞 𝘀𝘁𝗲𝗽𝘀/𝘀𝗲𝗰 🐡 + 🚗
youtu.be/LfQ324R-cbE?...
Excited to share 𝙋𝙪𝙛𝙛𝙚𝙧𝘿𝙧𝙞𝙫𝙚 2.0: A fast, friendly driving simulator with RL training via PufferLib at 𝟯𝟬𝟬𝗞 𝘀𝘁𝗲𝗽𝘀/𝘀𝗲𝗰 🐡 + 🚗
youtu.be/LfQ324R-cbE?...
PufferDrive 2.0 release
YouTube video by Daphne Cornelisse
youtu.be
December 30, 2025 at 4:12 PM
What if you could train agents on a 𝗱𝗲𝗰𝗮𝗱𝗲 of driving experience in 𝘂𝗻𝗱𝗲𝗿 𝗮𝗻 𝗵𝗼𝘂𝗿, on a single GPU?
Excited to share 𝙋𝙪𝙛𝙛𝙚𝙧𝘿𝙧𝙞𝙫𝙚 2.0: A fast, friendly driving simulator with RL training via PufferLib at 𝟯𝟬𝟬𝗞 𝘀𝘁𝗲𝗽𝘀/𝘀𝗲𝗰 🐡 + 🚗
youtu.be/LfQ324R-cbE?...
Excited to share 𝙋𝙪𝙛𝙛𝙚𝙧𝘿𝙧𝙞𝙫𝙚 2.0: A fast, friendly driving simulator with RL training via PufferLib at 𝟯𝟬𝟬𝗞 𝘀𝘁𝗲𝗽𝘀/𝘀𝗲𝗰 🐡 + 🚗
youtu.be/LfQ324R-cbE?...
I am not much into the prompt attacks for LLMs, however, this paper has a nice formalism to descritbe that
"Attacker LLM"
arxiv.org/abs/2512.20806
"Attacker LLM"
arxiv.org/abs/2512.20806
Safety Alignment of LMs via Non-cooperative Games
Ensuring the safety of language models (LMs) while maintaining their usefulness remains a critical challenge in AI alignment. Current approaches rely on sequential adversarial training: generating adv...
arxiv.org
December 30, 2025 at 12:16 AM
I am not much into the prompt attacks for LLMs, however, this paper has a nice formalism to descritbe that
"Attacker LLM"
arxiv.org/abs/2512.20806
"Attacker LLM"
arxiv.org/abs/2512.20806
Reacting
It is unavoidable that AI will be a major political issue soon. Or perhaps more appropriately: several major issues. I write more about this here:
togelius.blogspot.com/2025/12/maki...
togelius.blogspot.com/2025/12/maki...
Making AI Political
It is unavoidable that AI will be a major political issue soon. Or perhaps more appropriately: several major issues. As a technologist, I sy...
togelius.blogspot.com
December 30, 2025 at 12:11 AM
Reacting
Reposted
an argmin year in review
an argmin year in review
.
www.argmin.net
December 29, 2025 at 3:58 PM
an argmin year in review
Reposted
Chimera: State Space Models Beyond Sequences
Aakash Lahoti, Tanya Marwah, Ratish Puduppully, Albert Gu
Action editor: Hankook Lee
https://openreview.net/forum?id=yv0TUssepk
#embeddings #imagenet #attention
Aakash Lahoti, Tanya Marwah, Ratish Puduppully, Albert Gu
Action editor: Hankook Lee
https://openreview.net/forum?id=yv0TUssepk
#embeddings #imagenet #attention
December 29, 2025 at 5:19 AM
Chimera: State Space Models Beyond Sequences
Aakash Lahoti, Tanya Marwah, Ratish Puduppully, Albert Gu
Action editor: Hankook Lee
https://openreview.net/forum?id=yv0TUssepk
#embeddings #imagenet #attention
Aakash Lahoti, Tanya Marwah, Ratish Puduppully, Albert Gu
Action editor: Hankook Lee
https://openreview.net/forum?id=yv0TUssepk
#embeddings #imagenet #attention
add your own papers!
wanted to say that the year was excellent, a lot of positive steps into llm-based mas and improved convergence for imperfect info games
however, too much of adaptation too little of exploration
looking forward to 2026
wanted to say that the year was excellent, a lot of positive steps into llm-based mas and improved convergence for imperfect info games
however, too much of adaptation too little of exploration
looking forward to 2026
Multi-agent review of the year?
Bluesky exclusive!
Bluesky exclusive!
December 25, 2025 at 8:07 AM
add your own papers!
wanted to say that the year was excellent, a lot of positive steps into llm-based mas and improved convergence for imperfect info games
however, too much of adaptation too little of exploration
looking forward to 2026
wanted to say that the year was excellent, a lot of positive steps into llm-based mas and improved convergence for imperfect info games
however, too much of adaptation too little of exploration
looking forward to 2026
Reposted
Yue Lin, Shuhui Zhu, Wenhao Li, Ang Li, Dan Qiao, Pascal Poupart, Hongyuan Zha, Baoxiang Wang
Policy-Conditioned Policies for Multi-Agent Task Solving
https://arxiv.org/abs/2512.21024
Policy-Conditioned Policies for Multi-Agent Task Solving
https://arxiv.org/abs/2512.21024
December 25, 2025 at 5:48 AM
Yue Lin, Shuhui Zhu, Wenhao Li, Ang Li, Dan Qiao, Pascal Poupart, Hongyuan Zha, Baoxiang Wang
Policy-Conditioned Policies for Multi-Agent Task Solving
https://arxiv.org/abs/2512.21024
Policy-Conditioned Policies for Multi-Agent Task Solving
https://arxiv.org/abs/2512.21024
Multi-agent review of the year?
Bluesky exclusive!
Bluesky exclusive!
December 22, 2025 at 11:14 PM
Multi-agent review of the year?
Bluesky exclusive!
Bluesky exclusive!
slightly less iconic and maybe a bit flatter than Pendulum themselves, but still an excellent UK DnB from Metrik
soft and energising at the same time
www.youtube.com/watch?v=tL5E...
maybe good for a cardio sesh, @sharky6000.bsky.social
soft and energising at the same time
www.youtube.com/watch?v=tL5E...
maybe good for a cardio sesh, @sharky6000.bsky.social
Metrik - Synchronise
YouTube video by UKF Drum & Bass
www.youtube.com
December 20, 2025 at 10:11 PM
slightly less iconic and maybe a bit flatter than Pendulum themselves, but still an excellent UK DnB from Metrik
soft and energising at the same time
www.youtube.com/watch?v=tL5E...
maybe good for a cardio sesh, @sharky6000.bsky.social
soft and energising at the same time
www.youtube.com/watch?v=tL5E...
maybe good for a cardio sesh, @sharky6000.bsky.social
cool channel and videos
www.youtube.com/watch?v=4ac1...
www.youtube.com/watch?v=4ac1...
Using Evolution and Neural Networks to optimize Sailboat Races
YouTube video by Pezzza's Work
www.youtube.com
December 20, 2025 at 9:19 PM
cool channel and videos
www.youtube.com/watch?v=4ac1...
www.youtube.com/watch?v=4ac1...
Reposted
My colleague Sanjay Ghemawat & I have done a fair bit of performance tuning of various pieces of code. We wrote an internal Performance Hints document ~2 years ago as a way of identifying some general principles & we've recently published a version of it externally.
Doc: abseil.io/fast/hints.h...
Doc: abseil.io/fast/hints.h...
December 19, 2025 at 10:25 PM
My colleague Sanjay Ghemawat & I have done a fair bit of performance tuning of various pieces of code. We wrote an internal Performance Hints document ~2 years ago as a way of identifying some general principles & we've recently published a version of it externally.
Doc: abseil.io/fast/hints.h...
Doc: abseil.io/fast/hints.h...
still consider this paper an underrated banger btw
there was a couple of talks about it
one of them is www.youtube.com/watch?v=UKY0...
there was a couple of talks about it
one of them is www.youtube.com/watch?v=UKY0...
December 19, 2025 at 7:08 PM
still consider this paper an underrated banger btw
there was a couple of talks about it
one of them is www.youtube.com/watch?v=UKY0...
there was a couple of talks about it
one of them is www.youtube.com/watch?v=UKY0...
Reposted
🌍 Planet Wars: RTS AI Competition
→ simonlucas.github.io/planet-wars-...
🏃 LoRR: League of Robot Runners
→ Website coming soon!
📅 Mark your calendars and start preparing!
→ simonlucas.github.io/planet-wars-...
🏃 LoRR: League of Robot Runners
→ Website coming soon!
📅 Mark your calendars and start preparing!
December 19, 2025 at 6:35 PM
🌍 Planet Wars: RTS AI Competition
→ simonlucas.github.io/planet-wars-...
🏃 LoRR: League of Robot Runners
→ Website coming soon!
📅 Mark your calendars and start preparing!
→ simonlucas.github.io/planet-wars-...
🏃 LoRR: League of Robot Runners
→ Website coming soon!
📅 Mark your calendars and start preparing!
Democratisation of AI research like that, making complicated things available to everyone — that is sweet.
Do you use standard or batched variant of MCTS, though?
Do you use standard or batched variant of MCTS, though?
Hello! 👋
Are you interested in AI for board games using language models? Want to do some hobby tinkering with fine-tuning or RL?
We've released an easy-to-follow example colab that fine-tunes Gemma models via Kauldron to mimic an MCTS player.
Details here: github.com/google-deepm...
♟️🎲♦️♠️♥️♣️✨🎉
Are you interested in AI for board games using language models? Want to do some hobby tinkering with fine-tuning or RL?
We've released an easy-to-follow example colab that fine-tunes Gemma models via Kauldron to mimic an MCTS player.
Details here: github.com/google-deepm...
♟️🎲♦️♠️♥️♣️✨🎉
2025 Wrap-up: Fine-tuning Gemma with Kauldron Example ✦︎ · Issue #1414 · google-deepmind/open_spiel
Hello everyone! We've been hard at work this year working on OpenSpiel 2.0, which will be better than ever. Major developments have been underway to make working with language models easier. I'm lo...
github.com
December 19, 2025 at 6:42 PM
Democratisation of AI research like that, making complicated things available to everyone — that is sweet.
Do you use standard or batched variant of MCTS, though?
Do you use standard or batched variant of MCTS, though?
Reposted
Whether you're into robotics, multi-agent systems, or strategic AI, there's a challenge waiting for you:
⚓ MCTF: The Third International Maritime Capture the Flag Competition
→ www.mctf26.com
📊 NSGYM: Evaluating Decision Agents under Non-Stationarity
→ nsgym.io/aamas2026_co...
⚓ MCTF: The Third International Maritime Capture the Flag Competition
→ www.mctf26.com
📊 NSGYM: Evaluating Decision Agents under Non-Stationarity
→ nsgym.io/aamas2026_co...
December 19, 2025 at 5:51 PM
Whether you're into robotics, multi-agent systems, or strategic AI, there's a challenge waiting for you:
⚓ MCTF: The Third International Maritime Capture the Flag Competition
→ www.mctf26.com
📊 NSGYM: Evaluating Decision Agents under Non-Stationarity
→ nsgym.io/aamas2026_co...
⚓ MCTF: The Third International Maritime Capture the Flag Competition
→ www.mctf26.com
📊 NSGYM: Evaluating Decision Agents under Non-Stationarity
→ nsgym.io/aamas2026_co...
gimme snow
December 19, 2025 at 3:57 PM
gimme snow
cooperative foundation summer school lectures are live
www.youtube.com/watch?v=e0oi...
Seeing some fascinating ones
www.youtube.com/watch?v=e0oi...
Seeing some fascinating ones
Civilizational Resilience in the World of Artificial Intelligence by Nora Ammann
YouTube video by Cooperative AI Foundation
www.youtube.com
December 18, 2025 at 8:37 PM
cooperative foundation summer school lectures are live
www.youtube.com/watch?v=e0oi...
Seeing some fascinating ones
www.youtube.com/watch?v=e0oi...
Seeing some fascinating ones
whoa
🚨 Chelsea head coach Enzo Maresca is high among the candidates Manchester City are considering in case Pep Guardiola exits the Etihad Stadium next summer.
More from @david-ornstein.bsky.social
More from @david-ornstein.bsky.social
Chelsea’s Enzo Maresca high on Manchester City’s list in event of Pep Guardiola exit
City are advancing contingency planning as there is a growing anticipation of this being the Catalan manager's last season.
www.nytimes.com
December 18, 2025 at 2:04 PM
whoa
Reposted
Unlike board games, real-world strategic interactions are messy. Traditional game theory thus needs a boost for the age of agentic AI. Our #AAMAS2026 workshop "Strategic Engineering"(sites.google.com/view/se-aama...) in Cyprus aims to bridge the gap. Come join us to unlock truly strategic AI!
December 18, 2025 at 10:56 AM
Unlike board games, real-world strategic interactions are messy. Traditional game theory thus needs a boost for the age of agentic AI. Our #AAMAS2026 workshop "Strategic Engineering"(sites.google.com/view/se-aama...) in Cyprus aims to bridge the gap. Come join us to unlock truly strategic AI!
Published here: cyprusconferences.org/aamas2026/wo...
December 18, 2025 at 10:43 AM
Published here: cyprusconferences.org/aamas2026/wo...
Reposted
An earlier book I co-authored is on procedural content generation for games, the area of game AI that I have done the most research in. It's from 2016, but there's lots of useful stuff in there if you're into PCG:
www.pcgbook.com
www.pcgbook.com
Procedural Content Generation in Games book
www.pcgbook.com
November 25, 2024 at 5:04 AM
An earlier book I co-authored is on procedural content generation for games, the area of game AI that I have done the most research in. It's from 2016, but there's lots of useful stuff in there if you're into PCG:
www.pcgbook.com
www.pcgbook.com