Texel Support is live on every page of Texel, free for every visitor.
→ Docs: ask anything about Texel, and the answer comes from the live docs with the page to read next
→ Your wallet: once connected, Support reads your credits, instances, and outputs for you
→ Tickets: one message opens a ticket with the Texel team
→ Your keys: it never asks for a seed phrase or a private key
Answers stream in word by word, written by our fine-tune of Qwen3.5-9B with open weights on @huggingface.
We bought back 2.69M $TEXEL to fund Protocol Compute: the GPUs that run Texel's own AI.
The path, every step onchain:
→ Buyback: 2,686,035 $TEXEL off the market
→ Move: all of it into the Protocol Compute wallet of Texel
→ Credits: 686.32 Texel credits bought with that $TEXEL
Texel Support is the first service it pays for. Texel pays for its own GPU time the way every customer does, and as these credits get spent, the hourly burn takes their $TEXEL out of supply. That funds at least a month of Protocol Compute.
Every step on Etherscan:
Buyback: robin.etherscan.io/tx/0x5363…
Move: robin.etherscan.io/tx/0x86c6…
Credits: robin.etherscan.io/tx/0xf069…
1
4
426
Texel Support is in public testing on texel.run, and every promise of this post is kept. It runs on Texel with credits that $TEXEL bought, and every credit it uses sends its $TEXEL to the hourly burn, which has burned 24,062 $TEXEL since Nov 11.
→ Model: Qwen3.5-9B, trained on a Texel A40 rental with its receipt sealed onchain
→ Answers: it searches the docs before every reply, and got 77.8% of our test chats right against 62.5% for its base model
→ Your account: sign in, and it checks your balance, rentals, outputs, and receipts
→ Tickets: report a bug or a complaint, and it files a ticket for the team
→ Weights: public on @huggingface for anyone to run
→ Compute: serverless on Texel, with a GPU that starts when a question comes in
Texel is building its own AI agent for customer support. It will train on a GPU rented on Texel.
→ Base model: Qwen3.5-9B from @Alibaba_Qwen, 6x the size of our first support AI
→ Training: LoRA on a 48 GB NVIDIA GPU, with an onchain receipt for the whole run
→ Answers: drawn from the live Texel docs, so rates and the roadmap stay current
→ Tools: checks on your Texel credits, instances, outputs, and receipts once you sign in
→ Tickets: a direct line to the team for bug reports and complaints
Its weights will go public on @huggingface.
The same 48 GB GPU is yours from $0.41 an hour, 30% off at launch. Run your agent on Texel.
Every burn comes straight from the Texel credits contract. The newest one: robin.etherscan.io/tx/0x88f7…
Every $TEXEL that moves through the contract, burns included: robin.etherscan.io/token/0x0…
Press Support at the bottom right of any page, or read what it can do: texel.run/docs?v=2#support
4
234
Gm to everyone who keeps building while the market is down.
Texel is on three builds right now:
→ Texel Support, the AI agent we trained on an A40 rented on texel.run
→ Texel Serverless, your own model on NVIDIA GPUs that drop to zero when idle
→ $NVDA and more Robinhood Chain tokens to pay for credits
Texel Support's weights, open to everyone: huggingface.co/texellabs/tex…
Paying with any token is the next build on the Texel roadmap.
Pick a token from your wallet, pay, and your Texel credits are ready for every output and GPU rental.
→ Live today: $USDG, $ETH, and $TEXEL, all at 30% off
→ First to come: $NVDA, so tokenized NVIDIA stock pays for NVIDIA GPUs
→ After $NVDA: a few handpicked Robinhood Chain tokens
→ Once it's live: requests to list other Robinhood Chain tokens
Every token we add has a quality project behind it and deep liquidity, so each payment gets a fair price.
All 49 Texel ships since Nov 2 are live on @HeyResearch, and this one will join them.
Every receipt on Texel now sits one click from its proof onchain.
→ Press Receipt beside any output or rental
→ The Verify page matches it to Robinhood Chain in your browser
→ Check this hour rebuilds the whole hour and matches its root
The video walks through the A40 rental that trained our support AI: 13,681 GPU seconds for 2.2422 credits, sealed at 20:02 UTC.
Check any hour yourself, and see the A40's seal on Etherscan:
texel.run/receipts?v=2
robin.etherscan.io/tx/0x3ff4…
1
9
568
Texel Serverless is in development. Bring your own AI model, and NVIDIA GPUs will run it the moment a request comes in.
→ Models: from @huggingface or your own container
→ Scale: up with the requests, back to zero when they stop
→ Price: per second of GPU work, in Texel credits
→ Endpoint: one URL for your app or your agent
→ Receipts: one for every request, sealed onchain each hour
An idle model will cost nothing.
Buy credits with $TEXEL: texel.run/billing?v=2
The burns so far on Etherscan:
robin.etherscan.io/tx/0x5817…
robin.etherscan.io/tx/0x271b…
robin.etherscan.io/tx/0x372d…
5
14
603



