FYI, this to me points to an inference bug, bad sampling, or a non-native quant. OpenRouter is known to route requests to absolutely terrible, borked implementations. A model like DeepSeek V4 Flash shouldn't be making syntax errors like this.

"this is step one. bigger models are already in development with infrastructure scaling to match. private api preview open to select partners today, with plans to open-source future versions. incredibly proud of the MSL team. excited for what’s to come!"

https://x.com/alexandr_wang/status/2041909388852748717

1. Duly noted on USDC and other payment options - I have to see how easy this is do to as we're using stripe.

2. Teams and orgs are very high on the priority list and we hope to have something on this front very soon.

To keep up with our development, Discord is probably the best place: https://discord.gg/ptDRKRJpPV

The productivity gains unleashed by coding agents have made everything else an unexpected bottleneck. In an effort to make the most out of this new paradigm, we ceded a lot of ground in product thinking, thoughtful UX, and design excellence. In other words, the pace of tooling for deciding what to build has not kept up.

Mowgli is inspired by, in equal parts, Figma and Claude's plan mode. It evolves a detailed specification and designs for every screen and state of the product on an infinite canvas. Owing to this UI, it can quickly mock up new features and flows, redesign existing ones, and show you variations side by side. LLMs are amazing at helping you explore vast solution spaces, but most current tooling focuses on getting narrowly perfect output based on a well-defined spec. We try to bridge that gap with Mowgli.

When you're ready to build, Mowgli offers a .zip export with a SPEC.md and unopinionated design .tsx files for your screens, in a perfect format for coding agents. Or alternatively, full, pixel-perfect export of all screens to Figma.

In this early stage, we support two entrypoints: (1) building a product from scratch through a guided experience, or (2) importing an existing product from Figma. We have a powerful, almost pixel-perfect Figma to code + spec pipeline that works on files of any size - from 0 to 300+ frames.

We're working hard on other ways to get your existing products into Mowgli (and would appreciate hearing about what you would like!)

- Timelapse of making a functional second brain app in Mowgli + Claude Code: https://www.youtube.com/watch?v=HeOoy8WDmMA

- Sample project designed and specced entirely by Mowgli (demo, no login needed): https://app.mowgli.ai/projects/cmluzdfa0000v01p91l5r61e3?the...

- Sample Figma import of the whole Posthog product, ready for iteration: https://app.mowgli.ai/projects/cmkl0zqng000101lo8yn2gqvd?dem... (thanks PostHog for being radically open with public Figma files!)

Comments URL: https://news.ycombinator.com/item?id=47197662

Points: 6

# Comments: 4

Comments URL: https://news.ycombinator.com/item?id=46698829

Points: 3

# Comments: 0

[0]: https://arxiv.org/abs/2501.12948

You can get a bare metal AX162 from Hetzner for 200 EUR/mo, with 48 cores and 128GB of RAM. For 4:1 virtual:physical oversubscription, you could run 192 guests on such a machine, yielding a cost of 200/192 = 1.04 EUR/mo, and giving each guest a bit over 1GiB of RAM. Interestingly, that's not groundbreakingly cheaper than just getting one of Hetzner's virtual machines!

Irrlicht had its editor (irrEdit), a sound system (irrKlang), and some basic collision detection and FPS controller was built right into the engine. This was enough to get you a considerable way through a fully featured tech demo, at the very least. (I even remember Irrlicht including a beautiful first-person tech demo of traversing a large BSP-partitioned castle level.)

However, for those not afraid to stitch these additional parts from other promising libraries (or derive them from first principles, as was fashionable), OGRE offered more raw rendering prowess: a working deferred shading system (this was the heyday of deferred shading), a pop-less terrain implementation with texture splatting, and more impressive shader and rendering pipeline support, with the Cg multi-platform shading language. I remember a fairly impressive ocean surface and Fresnel refraction/reflection demos from OGRE at the time.

The story of 10x developers among us is not a myth... if anything, it's understated.

- The compute requirements would be massive compared to the rest of the industry

- Not a single large open source lab has trained anything over 32B dense in the recent past

- There is considerable crosstalk between researchers at large labs; notice how all of them seem to be going in similar directions all the time. If dense models of this size actually provided benefit compared to MoE, the info would've spread like wildfire.