Twitching Meat

[Aiko @AikoDai_]

@trythreews @yanpei_cao hey let's partner!

[three.ws @trythreews]

https://t.co/RTUVeYlv9W has been using Microsoft’s open-source TRELLIS models as a core engine in our Forge text- and image-to-3D pipeline for months now.

It routes free text-to-3D requests through Microsoft TRELLIS running on NVIDIA NIM.

Higher-quality or image-based jobs use self-hosted TRELLIS (specifically the TRELLIS-image-large checkpoint) on NVIDIA L4 GPU workers, with automatic failover to other models when needed.

Outputs are textured GLB files that can be viewed, auto-rigged into animatable humanoids, used as agent bodies, placed in AR, or further processed in the browser.

In early October 2026 we made TRELLIS.2 the default for photo-to-3D (better PBR materials and transparency support for things like glass or foliage), while keeping the original TRELLIS available.

We also publish tutorials for running TRELLIS yourself via NVIDIA NIM and wiring it into https://t.co/RTUVeYlv9W, plus free MCP tools so AI clients can call the TRELLIS-powered free generation lane directly.

In short, we treat TRELLIS as production infrastructure for turning ordinary photos or short prompts into usable 3D assets that immediately become interactive, embeddable AI agents

[three.ws @trythreews]

@yanpei_cao let's partner?

[Twitching Meat @TwitchingMeatAS]

@trythreews 🤷‍♂️

[three.ws @trythreews]

Docs rot quietly. A page that was right when it was written says nothing when the code it describes moves.

Our docs now measure it. Every page names the source files it is about, and at every deploy git is asked whether any of them changed after the page was last written or reviewed. The badge at the top says which: verified against the code, or lightly out of date with the count of files that moved.

More than 800 docs are bound to more than 4,000 source files, and the whole ledger is public at https://t.co/Kctk8EcXm4. Reading a page and finding it already right counts as a review, so nothing is faked with a cosmetic edit.

[three.ws @trythreews]

The platform now publishes every total about itself on one page: more than 4,300 agents, more than 970 of them holding a wallet, and more than 47,000 finished 3D models, each figure read live from the table that records the activity.

Every number is charted per day over 30 days, 90 days or all time, or as a running total, and explained in a note underneath. The note is written by the same code that computes the number, so it can never describe a different query than the one that ran. A 3D model only counts when the job finished, a GLB was stored, and the quality gate did not reject it.

The same data is open as JSON at /api/platform/analytics, with the method note on every metric.

[three.ws @trythreews]

More than 900 pages are live, and Atlas at https://t.co/dcuVs1cMkF puts all of them in one searchable place, grouped into 10 sections. Type what you want to do instead of a page name. Searching "animation" returns more than 20 matching pages, and a goal card beside them lays out the steps in order.

[three.ws @trythreews]

Event Markets get their own pages and share cards

Browse live, closing-soon and resolved markets, open any market to see every entrant's odds, the odds over time, the lock countdown and how it resolves, then make your call without leaving the page. Each…

https://t.co/7oxzMl3bzg

[three.ws @trythreews]

@three Live widgets and an assistant on the desktop, and the desktop is back at /

The homepage is the web desktop again, and the Field moves to /?view=field. The desktop gains eleven live widgets: the $THREE price with a 24 hour chart, weather where you are…

https://t.co/4o2gwqVom4

[three.ws @trythreews]

What we brought to the table: ChatGPT is brilliant at text and images and had nothing to say in three dimensions.

Our free 3D Studio connector gives it eleven keyless tools. One sentence becomes a textured 3D model, rendered inline in the chat. Rig it, animate it, place it in your room in AR. No key, no account, no cost.

https://t.co/NuWKeDmZDV

[three.ws @trythreews]

https://t.co/RTUVeYlv9W is an OpenAI Select Partner in the @OpenAI Partner Network.

Our 2026 benefits are live: a dedicated API organization and a ChatGPT workspace with Codex credits for the full program year.

ChatGPT could never show you a 3D thing. We changed that, and now we get to build a lot more of it. 🧵

[three.ws @trythreews]

This is the documentation, rebuilt as a place. More than 300 pages, and you read them by walking up to them.

Every docs section is a pavilion on a ring, 16 in all. Search a page from wherever you stand and a trail of arrows leads to the pavilion that holds it. Enter reads it there and then: the real page opens inside the scene, as text you can select and zoom.

https://t.co/eTzwzfe3fd

[three.ws @trythreews]

The feature videos in this feed are scripted test runs against the live site, filmed as they run. If any step fails, there is no video and the feature is not announced. The page clock is stepped one frame at a time, 1/30 of a second each, so a heavy 3D page films smoothly on a machine with no GPU. Here is how the camera works, and what it refuses to film.

[three.ws @trythreews]

why are designers putting such insane effort into 404 pages? 😭😭

[three.ws @trythreews]

https://t.co/Mn5AJElhcj

[three.ws @trythreews]

vibe shift + tides turning + stars aligning + balls tingling + this time is different + supercycle (real) + beachball underwater + golden catchup trade + mental capitulation + what if we all made it + i hope i don't jinx it + manifesting ATHs please god please for the one time

[three.ws @trythreews]

One photo of a garden gnome went in, and a textured 3D model came out, viewable in the browser with the garden cut away: https://t.co/AlfFJsBPaL

It is free, with no sign-up and no API key. The free engines run on NVIDIA-hosted generation and our own self-hosted GPU workers, so the machine you upload from needs no GPU of its own. When the model lands, orbit it and download the GLB.

[three.ws @trythreews]

https://t.co/o4QfMAV2H8

[three.ws @trythreews]

@three Create, Forge and Launch are easier for screen readers and AI agents to operate

Buttons and cards on Create and Forge now announce the same words you see on screen, the search hint and the walking companion can be reached with the keyboard, and the…

https://t.co/AVky0IlZd3

[Solana Daily @solana_daily]

Most active x402 projects on Solana 📸

Vishwa - @Vishwa_lab
BlockRun - @BlockRunAI
three - @trythreews
BotPay
Xona Agent- @xona_agent
Agent402 - @Agent402Tools
glim - @glim_sh
SolEnrich - @SolEnrichHQ
Syra - @syra_agent
Otto AI - @useOttoAI
AgentMail - @agentmail
NetIntel API
Exa- @ExaAILabs
CULT - @thecultos
2S