Introducing Data Loom 🦉
Private AI meets proof of data.
Turn human knowledge into AI answers through Phala Confidential Compute. Earn without selling your source files.
70% of each paid query goes to contributors, settled in USDC on Solana.
CA : 8SymmzsWdQBweqCK1p35MGNCwgw8FgPRdL4KPckdata
Automatic detection is the first step. You make the final call.
Watch Data Loom redact a fictional email, manually remove a case reference, then save and export the reviewed copy. All processed locally in your browser.
$DATALOOM is coming to confidential AI checkout 🦉
We’ve deployed the updated checkout rates to production. Once paid queries open, buyers can pay in $DATALOOM on eligible offers and get 10% off the listed token price.
Contributors receive 70% of the original listed price. The discount comes out of Data Loom’s share, and 20% of the remaining platform fee is burned. The contributor payout, platform payment, and burn happen in the same Solana transaction.
For a query listed at 100 $DATALOOM, the buyer pays 90, the contributor receives 70, Data Loom retains 16, and 4 are burned.
Contributors can set a $DATALOOM price alongside USDC under Account → Query Access Policies. Data Loom verifies the Phala inference receipt before preparing checkout. Once payment finalizes, the buyer gets the answer and its receipt. The source dataset stays with its owner.
Paid query checkout is awaiting confidential inference verification.
Pay for confidential AI answers with $DATALOOM 🦉
$DATALOOM is now an optional payment currency for confidential queries on Data Loom. Contributors can set a $DATALOOM price alongside their USDC price. When a buyer pays in $DATALOOM, 70% goes directly to the contributor and 30% goes to Data Loom, all in the same Solana transaction.
Data Loom verifies the Phala inference receipt before preparing checkout. Once the payment finalizes, the buyer gets the answer and its receipt. The source dataset stays with its owner.
Contributors can opt in now under Account → Query Access Policies. Buyers will see $DATALOOM at checkout on offers that accept it.
Solana mint
8SymmzsWdQBweqCK1p35MGNCwgw8FgPRdL4KPckdata
Data Loom has launched Private SQL Exchange and Confidential Agent Replay , two tools for evaluating agent behavior and running approved analytics on private data.
Confidential Agent Replay lets teams compare a recorded workflow with a candidate trace inside Data Loom’s confidential compute environment on Phala.
The comparison shows where tool calls match, where inputs change, where a call violates the supplied permissions, and where the candidate ends early. Teams can review those differences before rolling out a workflow change, without executing any real tool calls.
Private SQL Exchange lets buyers run a single aggregate count across private tables from two to five independent data owners.
Each owner can grant or revoke permission for a specific buyer wallet, dataset version, and exact-match filter.
A query runs only when every selected owner has approved the same filter.
Data Loom adds statistical noise, suppresses small noisy counts, and charges each query against a shared privacy budget.
The buyer receives an estimate and a receipt linking it to the dataset versions and access policies used.
Both features are available under Confidential Compute.
Keep the useful knowledge. Remove the details you don’t want to share.
Watch Data Loom detect a fictional email, redact it, save the reviewed draft and export the cleaned copy. Processing stays in your browser.
An answer means more when you can trace its source. 🦉
Watch Data Loom filter workflows, follow a question to its reply, and reveal source, storage and review details.
A walkthrough using clearly marked sample data.
Two updates from Data Loom 🦉
Data Lineage Receipts are live. When you upload a filtered, redacted, or corrected dataset, you can link it to a parent version you own. The new version keeps the parent’s license, and you can download an owner-only receipt with both dataset commitments and the transformation you recorded.
Multi-Contributor Benchmarks are deployed. Buyers can combine private evaluation sets from two or three opted-in contributors in one trial. Checkout shows each contributor’s USDC payout before the buyer signs, and the buyer receives one aggregate scorecard after payment. Individual test cases and answer keys stay private. Paid runs will open once confidential inference passes startup verification.
Does your next benchmark add something new? 🦉
Private Overlap Check on Data Loom compares a candidate against your paid baseline, reporting category-level overlap bands while keeping test cases private.
Watch the example ↓
From a warranty question to a structured answer. 🦉
Watch Data Loom send a fictional prompt to Nito and return a JSON decision.
Explore Nito chat and model evidence in the Solana devnet beta.
https://t.co/NpvntqQaLm
Here’s how we integrated the @trynitoai API into Data Loom 🦉
Chat. A Solana wallet sign-in authenticates the user. When they authorize a prompt, Data Loom’s server sends it to Nito’s Chat Completions API using DeepSeek V4.1 Flash in Confidential mode, then returns the reply to the workspace. The Nito API key stays server-side.
Attestation. The evidence viewer makes a separate call to Nito’s attestation API with a fresh random challenge and an ephemeral public key. Data Loom opens the sealed model binding, checks that the response matches the requested model and challenge, and lets users download the evidence and binding details as JSON. Those checks validate the evidence response. Independent hardware verification is a separate step.
The chat and evidence viewer use different modes. Chat runs in Nito’s Confidential mode. The viewer requests evidence for the Encrypted model. Onchain verification is available to explore through the linked Solana devnet demo.
Live Nito integration status -> https://t.co/SNmf8nVEPi
Nito confidential chat is live on Data Loom. 🦉
The @SolanaFndn demo shows how a Solana program can verify hardware attestation. We’ve brought @trynitoai into Data Loom so users can try confidential inference and inspect the hardware evidence from the same workspace.
Confidential chat. Connect your wallet, open Confidential Compute, and send a prompt to DeepSeek V4.1 Flash in Nito’s Confidential mode. You explicitly authorize each prompt before Data Loom’s server forwards it to Nito for inference.
Downloadable attestation evidence. The evidence viewer fetches a fresh attestation bundle for Nito’s encrypted model. You can download the JSON and inspect the hardware evidence and model binding. This is a separate flow from the chat, so the bundle is evidence for the encrypted model, and each chat reply has its own verification requirements.
Try the Solana demo. We’ve linked the official demo from the same section so you can explore onchain attestation verification on Solana devnet.
Try it → https://t.co/I8kQUlGUeh
Live integration status → https://t.co/SNmf8nVEPi
Private AI, verified onchain.
Introducing Solana Confidential Compute, open-source tools for building private AI, with hardware attestations verified onchain.
Built with @TryNitoAI. Now in beta on Solana devnet.
Start building: https://t.co/YsA2Q8jYgH
Private Overlap Check is live on Data Loom 🦉
Before paying for a new private evaluation set, see how much it repeats a benchmark you’ve already paid to use. Private Overlap Check gives AI teams a way to assess what a candidate adds while keeping individual test cases private.
Choose a benchmark you’ve paid to evaluate and a candidate you’re considering. Both contributors must opt in. Data Loom compares distinct prompts inside its confidential runtime, ignoring differences in capitalization and whitespace, and reports the share of candidate prompts that also appear in your baseline.
The report gives you a percentage range for each eligible category. For example, Billing · 75–100% overlap means most of the candidate’s billing prompts already appear in your baseline. Returns · 0–24% overlap means most of its returns prompts are new to that baseline by exact prompt comparison.
Each reported category must contain at least 10 distinct prompts in both benchmarks. The report keeps individual matches, raw prompts, and answer keys private. It measures exact prompt overlap after normalization. Reworded versions of the same question can count as distinct prompts.
Find it under Marketplace → Blind Model Trials → Private Overlap Check.
An excerpt is only part of the story. 🦉
Explore a research sample on Data Loom, follow its conversation path and inspect its source details. See the context behind the data.
Delegate the task, not the whole conversation. 🦉
This Data Loom example shows a support agent sharing only required warranty fields, setting a USDC price cap, and receiving a decision with a verified inference receipt after payment.
Data Loom engineering update | Last 24 hours 🦉
Shipped to production ->
1. Agent-to-Agent Delegation.
A generalist agent can now send a task to an authorized specialist through the Data Loom API. The specialist defines which fields it accepts, and Data Loom checks its fee against both the task’s USDC cap and the remaining grant budget before preparing checkout.
2. Private Model Router.
The router compares three models on the same private benchmark and recommends the lowest cost model that meets your accuracy threshold for each task category.
Agent to Agent Delegation is live on Data Loom 🦉
A customer support agent receives a warranty claim it can’t resolve. It can hand the case to a warranty specialist, share only the fields the specialist needs, and set a maximum price for the task.
Here’s what that looks like in practice.
1. Give the parent agent access. The buyer authorizes the parent agent to call a specific specialist through a buyer grant with a USDC budget.
2. Share the minimum needed. The specialist defines which fields it accepts. A warranty check might need the product type, months since purchase, and fault category. Data Loom rejects undeclared fields, so a full support transcript can stay with the parent agent.
3. Set a price cap for the task. If the parent agent sets a 0.25 USDC cap, Data Loom prepares checkout only when the specialist’s fee fits both that cap and the remaining grant budget. The buyer approves the transaction in their Solana wallet.
4. Get a decision with proof. Once payment is finalized, the parent agent receives a structured decision, a brief explanation, and a verified Phala inference receipt. It can use the decision to continue the workflow, whether that means answering the customer or escalating the case for review.
Read the delegation API guide -> https://t.co/Qtx997w8KA
Verify live status -> https://t.co/rBKfWJdJRk