A T

[A T @iDramedy]

@honx @awesomekling Hence, must be vigilant w our ability to focus our attention towards the stuff that matters.Tony Stark has to call on to Banner or Mr.Fantastic for some problems. Specialization is okay. Don’t try to eat all the cake. Internet doom scroll and Brainrot didn’t happen by accident.

[Patrick Smith @patrickgwsmith]

I’m working on a WebAssembly debugger. You can even time-travel backwards!

[A T @iDramedy]

@NotTuxedoSam @0xblacklight @humanlayer_dev Chicken and egg problem then?

[elvis @omarsar0]

If you maintain a hand-built agent harness, this one is worth your time.

(bookmark it)

I feel like everyone is sleeping on the idea of dynamically generating agent harnesses on the fly.

As you aim to own your harness, this is a topic more devs will lean into. Here is a great report discussing this topic.

JIT-Agent is a model whose output is an agent harness.

It formalizes the harness as a composable artifact under a fixed four-module protocol covering memory, planning, action protocol, and tool orchestration, then synthesizes one on the fly for any off-the-shelf agentic LLM.

It also repairs harnesses mid-execution and self-evolves by distilling performance signals from an expanding archive of prior configurations.

With JIT-Agent attached, DeepSeek-V4-Flash surpasses GPT-5.6 on DeepSearchQA (+9.1) and OdysseyBench (+4.3). GLM-5.2 gains up to +20.2 points.

The generated harnesses are also performance-competitive with mature runtimes like OpenCode and Claude Code.

Paper: https://t.co/RYD5Qbjfv6

Chat with Paper: https://t.co/RXnz7FnJqR

[A T @iDramedy]

@iamsausii @paulg That’s like saying “just make a faster compression algorithm”. Ok, but, what does the person know about compression in the first place? The history of it? What are the failures of the past and why what has stood the test of time endured?

[sunil pai @threepointone]

“Look at my incredible new factory!”

Yo that’s cool, what do you make?

“It’s highly optimised, fully automated, zero tolerance for defects and with a continuous feedback cycle”

Cool cool, so what do you actually make?

“I can interact with it on my phone, laptop, messenger, completely async, and the shared context means it’s always learning how to get better”

Very impressive, but what do you make?

“Every agent has full context, can spawn other agents, review their work, fix defects, and ship continuously.”

yes yes. WHAT DOES IT MAKE?

“Software.”

Oh nice. What software?

“Well right now we’re mostly using it to improve the factory.”

Improve it to make what?

“Anything!”

Such as?

“…a better factory.”

[A T @iDramedy]

@rauchg Come on @rauchg, let’s be real here and tell the truth. What you really mean to say is that you want to be one of the first investors of whichever person/company builds WestWorld.

[Sergey Levine @svlevine]

Latest Deep RL class lectures are now online!
https://t.co/GvqI1v3hgD
Thanks to @seohong_park, we now have CS185/285 for spring 2026 available to everyone to watch.

Course website here: https://t.co/U16rasTOAo

Apologies for a few recording glitches (it's not a perfect system).

[Valeriy M., PhD, MBA, CQF @predict_addict]

An absolute banger of a paper.

"A Gentle Introduction to Matrix Calculus" by econometrics legend Jan Magnus — one of the clearest explanations of matrix derivatives ever written. Published in the Journal of Econometrics in 2024.

If you work in econometrics, machine learning, statistics or optimisation, this paper is pure gold.

Free, in my Awesome Math Books list (econometrics section)

https://t.co/soOYrEQ2he

[Tariq @Grgo6_]

bare-metal rust across esp and pico, running in the browser. still can't quite believe this works

[Philippe Aghion @Ph_Aghion]

My Nobel lecture and the video of the Stockholm lecture are online.

It covers what creative destruction tells us about secular stagnation, firm dynamics, the middle-income trap, and AI as the next GPT revolution.

Paper: https://t.co/Viem3OXkiC
Video: https://t.co/9rfqTvWXMy

[A T @iDramedy]

@ege02 @daniel_mac8 lol, this whole industry is going to fuck around just to come right back to “yeah, it’s code. We need to use tight symbolic well defined machine checkable written instructions. Otherwise; what are we doing? money doesn’t grow on trees and we have yet to discover infinite energy”

[KIRI Engine - 3D Scanner App @KIRI_Engine_App]

We made a small open source project with 3D Gaussian Splatting.

KIRI-Maker lets you turn your 3DGS scans into something more than static captures:

✨ Particle effects
🎥 Camera path animations
🖐️ Gesture controls
🎬 Exportable videos

Explore, animate, and play with your 3D scans.

[Sida Peng @pengsida]

InfiniSplat (SIGGRAPH Asia 2026)​ is trending on @HuggingFace!
🚀 Turn a single image into a navigable 3DGS scene in <1 second.
🏠 Stunning indoor view synthesis quality.
👉 Try it now: https://t.co/AIAUaSC8tA
🌐 Project: https://t.co/O7jEBCR2wY

[InSpatio｜影溯 @InSpatio_AI]

2D is what the camera sees.👀
3D is what the world is. 🌍

Today, we’re proudly open-sourcing QuerySplat — an open-source feed-forward 3D model that turns a few unposed images into a navigable 3D scene in seconds (run on 4090🪶).

Your input: a few unposed images
Our output: a sharp, navigable 3D scene — generated in seconds! 🚀

🧠 organize scenes with 3D queries, not pixel-bound Gaussians
📷 no camera poses or manual calibration required
⚡ reconstruct new scenes in a single forward pass
✨ preserve sharp geometry and high-frequency appearance
🏆 achieve state-of-the-art results on DL3DV-Evaluation
🔓 code and model weights are now open-source
From pixels to scenes, we are making 3D generation scalable.

Try our free iOS app — it’s a lot of fun! 😆
🔗 https://t.co/VaguGSnRqZ
Try it on the Web: 🔗 https://t.co/3ZrCuGZCKS
Paper:
🔗 https://t.co/gMHB9b5vjR
Code & weights:
🔗 https://t.co/ofbNZMjtcF
Project page:
🔗 https://t.co/GRkQm4oTIh

#AIGC #3DGS #AI #Computervision #3D #artist #3dartist #3d

[A T @iDramedy]

@AlexJonesax Hey, it is called cernvmfs. You are welcome.

[Andrew Koh @andrewjkoh]

New paper w/ Drew on R&D races in the shadow of ruin: https://t.co/IeObR9LXUd.

There've been recent calls for slowdowns/pauses. A common objection goes: "if we slow down, others won't, so we shouldn't." What to make of this? We work out the game theory of how the frontier is shaped by:

- Competition: what do I gain from being ahead? lose from being behind?
- Coordination: when the tech becomes really dangerous I’d like to stop, but only if my rival also stops
- Transparency: how quickly are actions observed?
- Trust: how confident am I that my rival is rational?

[A T @iDramedy]

@EpsilonRho @ryan_t_lowe @confusionm8trix It follows that you must now seek bigger problems to solve. Think in terms of hybrid collective intelligence abundance. You want pods of humans solving bigger and bigger problems.

[A T @iDramedy]

@JazzyBobbalu @rfleury @sama Aye, Bobby. Let this one go. This is not the hill. It is a simple concept: be very present and involved with your kids even when it is or feels inconvenient. How do you think we ended up with the snowflake iPad/Screen kids problems. It is a slippery slope.

[Sasha Malysheva @aimalysheva]

Three books I can recommend for getting into multi-agent systems:

1/ "Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations", Shoham & Leyton-Brown (it's digital version is free at https://t.co/IQtoIteHsA)

2/ "An Introduction to MultiAgent Systems", Wooldridge (best first read, as it will help you to build the vocabulary)

3/ "Algorithmic Game Theory", Nisan, Roughgarden, Tardos, Vazirani (if you want the math behind why agents behave the way they do)

funny how a literature written for auctions and negotiating robots became required reading for people wiring LLMs together -- the incentive problems carried straight over, only the agents are new 😁