Ed Truckell

overheard a new insult: you have a short context window 💀

BBC Two - 2000 - UK

🚨Do you understand the irony of what just happened.. the FBI spent two years investigating Hillary Clinton for using a personal email for government work.. it dominated every news cycle.. it decided an election.. "lock her up.." the man they put in charge of the FBI.. just got his personal email hacked by Iran.. the DOJ confirmed it.. Iran-linked group "Handala" breached FBI Director Kash Patel's personal email and got his data.. think about the timing.. America has fired 850 Tomahawk missiles at Iran this month.. spent billions.. and Iran didn't fire back with a missile.. they walked through the front door of the FBI.. and here's what nobody wants to say out loud.. the FBI Director was using a personal email in the first place.. the same thing the FBI called a national security threat when Hillary did it.. the agency that decides what's a crime.. doesn't follow its own rules.. The rules only apply until you're the one making them.

Confidence-Based Mesh Extraction from 3D Gaussians Contributions: – We introduce a self-supervised confidence framework to 3D Gaussian Splatting, along with densification adaptations, improving surface reconstruction – We demonstrate how reducing variance in color- and normal blending resolves visual and geometric ambiguities, thereby improving the accuracy of surfaces – Motivated by an analysis of photometric losses in 3DGS, we propose a novel decoupled appearance embedding, improving reconstruction in challenging real-world datasets

Can you imagine how fast depth estimation has improved! InfiniDepth (github.com/zju3dv/InfiniD…) was accepted in CVPR 2026 and opensourced now. We used tsdf to create mesh directly to examine and you can see how clean the surfaces are. No refinement required at all. It is crazy. Congrats @pengsida #CVPR2026

Everyone is talking about DLSS5. And it actually works! (I'd love to give credit, but I don't know who made this. Wasn't me!)

When I was consulting for @HBO Silicon Valley, zero-loss compression was the holy grail Richard Hendricks chases that perfect middle-out algo could shrink everything w/out breaking a single bit. Google just did something even more practical for the AI era: TurboQuant compresses LLM key-value caches down to 3 bits per value using random orthogonal rotation + PolarQuant scalar quantization & optional 1-bit QJL residual correction. =>> 6× memory reduction, up to 8× faster attention (on H100), & 0 degradation on LongBench, Needle-in-a-Haystack, and RULER for models like Gemma. No retraining, no calibration needed. Fiction just got out-engineered by reality. 😅💚💚

One of the greatest pieces of media ever created

Absolute bombshell. Congressman Massie reveals that top law enforcement officer Pam Bondi was initially ready to release the Epstein files, but someone got to her and ordered her to cover it up. The deep state is actively protecting the elite pedophile ring.

It won't, but JMail should be winning every research prize and grant for the next five years minimum.

RIP Sora, you gave us the greatest ai video of all time

'My iPhone containing messages with my friend Peter Mandelson was stolen by a slim gentleman on a pedal bike. To save British taxpayers' money I turned off the free 5GB iCloud text messages backup. I now regard this as the end of the matter'

EXCLUSIVE: I’ve spoken to MPs, Labour Party whistleblowers, and a former frontbencher with DIRECT knowledge of Morgan McSweeney. We are now witnessing the start of a blatant government cover-up. Is McSweeney lying about the “stolen” phone? Here’s what we know so far:🧵

the judi dench cameo fucking sent me 🤣😭

In Breaking Bad, when Hector Salamanca is taken to the nursing home, a framed family photograph is briefly visible; it shows Hector seated alongside several of his nephews when they were children. At first glance, it looks like just another photo, but upon closer inspection, a very interesting detail emerges. The photograph features an older boy wearing a belt with a large buckle a highly distinctive style that, years later, we would see on Lalo Salamanca in Better Call Saul. This has led many fans of the series to speculate that this boy might be a young Lalo. If true, this would mean that Lalo had already made an indirect appearance in Breaking Bad long before the character was officially introduced in Better Call Saul. If this theory holds true, it stands as one of the most intriguing details demonstrating just how incredibly interconnected the entire universe created by Vince Gilligan truly is revealing that characters who seemed not to exist were actually there all along, right from the very beginning, without us ever realizing it.

Matt Berry + Star Trek = perfection (amazing edit by @Picardlooking)

🍺 LagerNVS (CVPR 2026) 🍺 LagerNVS is a generalizable, feed-forward, real-time Novel View Synthesis network which - performs rendering in real time, - generalizes to in-the-wild data, - works with and without known source cameras, - sets a new state-of-the-art among deterministic methods, - can be paired with a diffusion decoder for generative extrapolation. LagerNVS shows that 3D biases are useful for Novel View Synthesis but explicit 3D representations are not required to achieve them. We use 3D biases in (1) architecture design and (2) pre-training: (1) In NVS with explicit 3D representations (3DGS, NeRF) reconstruction is typically difficult and slow, but rendering is much faster and simpler. We mimic this process in the network design: we use a large (1B params) encoder and a small, lightweight decoder (ViT-B). This allows increasing the network capacity while still achieving real-time rendering. (2) The encoder, initialized from VGGT, was pre-trained with 3D reconstruction objectives, making the initial features 3D aware. Both substantially improve performance. Project page: szymanowiczs.github.io/lagernvs Code: github.com/facebookresear… Paper: arxiv.org/abs/2603.20176 Models: huggingface.co/collections/fa… Work done with @jianyuan_wang @MinghaoChen23 Christian Rupprecht and Andrea Vedaldi

If you've got an opinion you want on here - submit it to the form: forms.gle/1gNMupaNHQf8xe…

You know what a wave at the beach looks like. The water goes up and down, but the wave itself moves sideways toward the shore. The motion of the water is perpendicular to the direction the wave travels. That's a transverse wave. Light works the same way. The electric field wiggles up and down. The magnetic field wiggles sideways. Both perpendicular to the direction the light travels. Every electromagnetic wave you've ever encountered (WiFi, radio, X-rays, the infrared heat from your coffee) does this. Electric field one way, magnetic field the other way, energy goes forward. That's all your textbook teaches. Now imagine a slinky. You push one end. A compression travels along the slinky. The coils bunch up, then spread out, then bunch up again. The motion of the coils is in the SAME direction the wave travels. That's a longitudinal wave. Sound works like this. Air molecules compress and decompress along the direction the sound travels. Here's the thing nobody told you: Maxwell's original equations allow for an electromagnetic version of the slinky wave. An electromagnetic wave where the electric field oscillates ALONG the direction of travel, not perpendicular to it. And this wave has no magnetic field at all. Why does that matter? Because every technology that blocks electromagnetic waves (Faraday cages, metal shielding, the walls of a submarine) works by the same mechanism: the magnetic field of the wave induces electric currents in the conductor. Those currents resist the wave. That's the skin effect. That's why your WiFi dies behind a thick wall. A wave with no magnetic field doesn't trigger that mechanism. No magnetic field means no induced currents. No induced currents means no skin effect. The barrier that blocks your WiFi doesn't interact with this wave type at all. It passes through. Not because it's stronger. Because the blocking mechanism doesn't apply. Heaviside's simplification in 1884 removed this wave type from the equations. Not because anyone tested it and found it wasn't real. Because it complicated the math, and the simplified version was good enough for radio and light. 140 years later, four independent groups put it back. Same wave. Same math. Different formalisms. None cited each other. The wave was in the equations the whole time. Someone simplified it away. That's the paper. That's the whole argument. The textbook uses 6 of 16 electromagnetic components. The other 10 describe this wave, and more.

I need an update on this.