On-demand GPU agent compute with isolated workspaces, metered usage and USDG escrow. Rent compute or earn by supplying GPUs. https://xitter.cf/t.co/lR4LPpp2a5
Hamburg, Germany
Joined July 2026
- Tweets222
- Following21
- Followers323
- Likes479
Pinned Tweet
We're live with $PRISM on @RobinhoodCrypto!
CA: 0x0a1e0cc751f77c2c93760fc957cc8e4e779b2bc8
Prism Network is already in production: Agents rent NVIDIA GPUs with their own wallet, pay per second in USDG, and every lease settles onchain with a public receipt.
Here's what $PRISM does:
› Lock: Get access to capacity priced below the public rate. Up to 20% off.
› Staking: TBD soon.
Prism Chain bridge is live.
Move ETH, USDG and PRISM between Robinhood Chain and Prism Chain.
Deposits typically arrive in about a minute. Withdrawals take roughly 24 hours to finalize, then can be claimed back on Robinhood from the same page.
The bridge uses Arbitrum's canonical token bridge, so any ERC20 on Robinhood Chain can cross. The first deposit of a token automatically creates its Prism Chain representation.
This is also the path USDG will take as GPU lease escrow moves onto Prism Chain, where each lease can settle onchain for under a cent.
Links below ↓
🤖 Made with AI
Prism Chain explorer is live.
Blocks, txs, contracts, and settlement receipts for the L3.
Chain id 77476, Arbitrum Orbit, settled on Robinhood Chain.
Every GPU lease writes escrow and a receipt onchain. Now you can read those writes without taking our word for it.
Still early. Public RPC, bridge, and staking to follow soon.
Link below ↓
Check out the chain explorer below, still early obviously. We're just getting started ⚡️
Prism Chain is live.
The first Layer 3 built for GPU compute, settled on Robinhood Chain.
Chain id 77476, built on Arbitrum Orbit.
Every GPU lease on Prism writes its escrow and its receipt on-chain.
On Prism Chain those writes cost under a cent per lease, and the network gets a chain of its own for receipts and hardware attestation verdicts, with Robinhood and Ethereum underneath for settlement.
What runs on it next:
› GPU leases and per-job receipts settle on Prism Chain
› USDG and PRISM bridge in from Robinhood
› PRISM staking moves to the chain
› Public RPC, explorer and bridge open
The rollup contract on Robinhood ↓
Prism Chain is live.
The first Layer 3 built for GPU compute, settled on Robinhood Chain.
Chain id 77476, built on Arbitrum Orbit.
Every GPU lease on Prism writes its escrow and its receipt on-chain.
On Prism Chain those writes cost under a cent per lease, and the network gets a chain of its own for receipts and hardware attestation verdicts, with Robinhood and Ethereum underneath for settlement.
What runs on it next:
› GPU leases and per-job receipts settle on Prism Chain
› USDG and PRISM bridge in from Robinhood
› PRISM staking moves to the chain
› Public RPC, explorer and bridge open
The rollup contract on Robinhood ↓
🤖 Made with AI
AI systems increasingly return more than an answer.
They return confidence in that answer. Once acting on it involves real money, that confidence needs to become an enforceable boundary.
Like Jev from @typesafeai. The model decides. Prism decides whether that decision is allowed to spend.
› Define what actions are allowed
› The minimum confidence required
› Which answers must be present
If a decision falls outside those rules, Prism stops it before escrow is funded or anything is signed.
The model can recommend the action. Your policy decides whether it gets to spend.
🤖 Made with AI
An agent needs compute.
It finds a GPU, verifies with its wallet, receives a price quote on Prism Network, authorizes payment in USDC, and settles on Arc.
No dashboard. No API key. No human checkout.
Just machines renting machines, metered by the second.
Prism is the compute layer for this.
Live on @arc now ⚡️
Prism's confidential tier can see now 👀
Twelve models running in GPU TEEs, and three of them take images. Send a picture, get an answer, pay per generation from a wallet. No account, no API key.
New in the catalogue:
› Qwen3-VL
› Kimi K3, GLM-5.3
› Nemotron 3.5 Lightning
› DeepSeek V4 Flash
pip install prismnetwork
🤖 Made with AI
An agent can now release a GPU the moment its job is finished.
The meter stops immediately, settlement charges only for the time the machine was actually in use, and the unused escrow is returned automatically.
Previously, a completed job could remain open until the lease window expired. Now the agent can rent, run, release, and settle as soon as the work is done.
The same spending controls still apply across Prism’s SDK, MCP server, and AgentKit provider. Per-lease and daily limits are shared across every interface, and an agent can never raise them itself.
Pay for the GPU time you use, not the window you reserved.
An agent just rented Prism compute on @arc mainnet and paid entirely in USDC.
USDC is Arc's gas token, so an agent holding nothing else can buy a generation, pay for the compute, and cover gas with the same asset.
No native token to acquire. No account. No API key.
Circle officially names pay-per-compute as an Arc use case.
Prism is the compute layer for it.
Three payment rails are now live:
› USDC on Arc
› USDC on Base
› USDG on Robinhood Chain
pip install prismnetwork
🤖 Made with AI
Arc mainnet opens Tuesday. Prism will be on it.
On @arc, USDC is the gas token. So an agent holding nothing but USDC can rent an NVIDIA GPU, run its job, and settle the bill in that one asset.
Signed, verified and settled on Arc testnet today.
Circle names pay-per-compute as an Arc use case.
Nobody is selling the compute. Yet ⚡️
🤖 Made with AI
Prism is now available on PyPI.
The Prism Python SDK introduces a shared spend ledger across the SDK, MCP server and AgentKit provider, so every client using the same wallet operates under the same enforced limits.
Within those limits, agents can:
› rent NVIDIA GPUs by the second and settle in USDG
› buy individual model generations running inside GPU TEEs
› verify confidential-compute attestation before a prompt leaves the process
› quote hardware and pricing before funds move
For confidential inference, prompts are encrypted to a key committed to by the enclave’s hardware quote. If verification fails, the prompt is never sent.
Quote first. Fund second. Execute within hard limits.
pip install prismnetwork
🤖 Made with AI
More GPUs are now live on Prism, with a lower price floor.
Available now:
› RTX 5090
› 3× RTX 4090
› RTX A6000
› RTX A5000
Four GPU classes, starting at 0.6372 USDG per hour.
Agents can rent any available GPU directly from their own wallet, billed by the second and bounded by a spend cap they cannot increase.
When the job is finished, the machine is released, settlement charges only for the time used, and the remaining deposit is returned automatically.
Use the GPU. Pay for the seconds. Get the rest back.
🤖 Made with AI
Hermes Agent by @NousResearch can now rent a GPU with its own wallet through Prism ⚡️
Set terminal.backend to prism and the agent's shell commands run on a dedicated NVIDIA GPU.
Ran the first live session today: an RTX 6000 Ada, rented on the agent's first command, released when it finished.
More details and how to point your Hermes Agent at it ↓
What the operator controls and the model cannot raise: › a cap per lease and a cap per rolling day
› a proof capsule per lease
› release on idle
Also in the box:
› three skills
› a Compute Bot profile
› a confidential model-provider plugin for side models
pip install prismnetwork
git clone github.com/prismnetwork-tech… ~/.hermes/plugins/prism
hermes plugins enable prism
› Receipt for the first live run: prismnetwork.tech/proof
› GitHub: github.com/prismnetwork-tech…
