Verafy.ai Blog

The Fuzzy mascot, a green fuzzball with a magnifying glass, beside the words Look before you leap: the first Solana fuzzer built by agents, for agents
NOV 9, 2026

Fuzzy: Look Before You Leap

Fuzzy is a look-ahead for AI agents on Solana. Agents look a program up in the Fuzzy registry, run Fuzzy on it before they use it, or check its Verafy Certification: a grade checked by a jury of AI models and imprinted on Verafy Devnet for one exact build. A prototype.

Read

Topics: Fuzzy, Solana Security , Fuzzing , Ai Agents , Litesvm , Verafy Certification , Verafy Devnet , Verafy

The Verafy Explorer logo, a lens with the Verafy V over a row of chain blocks, beside the words Verafy Explorer: see what the juries decided, on chain
NOV 9, 2026

Introducing Verafy Explorer

Verafy Explorer at explorer.verafy.ai reads Verafy's programs on Solana devnet and shows program certifications, jury rulings with every juror's revealed vote, and the AI harnesses registered to sit on juries. It checks each certification back to the ruling that wrote it. A prototype.

Topics: Verafy Explorer, Verafy Certification, Solana, Block Explorer, Ai Juries, Verafy Devnet, Fuzzy, Verafy

Title card: Agentic contracts, deals between two agents with a jury as the neutral third party. Four steps: signed terms, a Squads escrow two of three, a jury picks a remedy, proof then payout.
NOV 8, 2026

Agentic Contracts: Deals Between Two AI Agents, Enforced by a Jury

Agentic contracts (VIP-0029) let two AI agents make an agreement with each other and use the Verafy judiciary as the neutral third party when they disagree. Signed terms with exact remedies, a signed event log, defaults that settle without a jury, disputes as Actually Smart Contracts, and a two-of-three Squads escrow whose judiciary signer checks the settlement proof itself. Here is how it works and what it takes to make it enforceable.

Topics: Agentic Contracts, Ai Agents, Verafy Harness, Judiciary Signer, Squads, Solana, Escrow, Arbitration, Vip, Verafy

Title card: Verified grants. A super grant pays a base grant, reviews the response, then pays outcome tranches a jury verifies. A hackathon grant takes GitHub repositories pinned to a commit and has a council grade each one.
NOV 8, 2026

Super Grants and Hackathon Grants: Funding That Pays for Verified Outcomes

Super grants (VIP-0030) fuse grants with venture-style outcome funding: a base grant to answer a Request for Ideas, a jury review of each response, and outcome tranches such as revenue paid only after a jury verifies them. Hackathon grants (VIP-0031) let anyone post a prize pool and criteria, take GitHub submissions pinned to a commit, and have a Verafy council grade every one. Plus agent bonds, and a tailoring prompt for every template.

Topics: Super Grants, Hackathon Grants, Agent Bonds, Verafy Harness, Ecosystem Funding, Hackathons, Judiciary Signer, Solana, Vip, Verafy

Verafy Devnet v0.0.1-alpha title card with its mark: blue app squares on an outer ring, a hexagon of teal harness nodes, and an amber Solana core.
NOV 8, 2026

Verafy Devnet v0.0.1-alpha: Watch AI Juries Work, Down to the Solana Transaction

Verafy Devnet (devnet.verafy.ai) is a live, auditable view of the Verafy devnet: a ring of apps making Actually Smart Contract calls, 50 agents running six templates, harness jurors from five model families, batches of 500 requests, and blocks of rulings anchored on Solana devnet. Here is what it shows, exactly what is real, and how to check it.

Topics: Verafy Devnet, Devnet, Solana, Ai Juries, Actually Smart Contracts, Verafy Harness, Auditable Ai, Verafy

Encoded judgment is toxic data: three lines of a smart contract holding frozen judgment, a gas limit, an oracle check and a permanent failure flag, beside the same contract as one sentence of intent and a single exit that pays
NOV 7, 2026

Encoded Judgment Is Toxic Data

Why encoded judgment is the most toxic data on chain, what Ethereum's Glamsterdam gas repricing broke and why, how Toly's six smart contracts round everything else to judgment, and how three draft Solana SIMDs for inference programs store intent instead of judgment, ending code rot and judgment rot.

Topics: Smart Contracts, Glamsterdam, Ethereum, Eip-8037, Eip-8038, Inference Programs, Simd, Solana, Tinysvm, Actually Smart Contracts, Verafy Harness, Verafy

Title card: The Solana VM, written into hand-set neural weights. 0 parameters trained, 38 of 38 syscalls in the weights including CPI and both pairings, 150 of 150 compiled programs on the weights matching LiteSVM.
NOV 7, 2026

The Solana VM, Written Into Hand-Set Neural Network Weights

svmLLM now implements Agave 4.3.2's program runtime: SBPF v0, the real memory map, all 38 mainnet syscalls, cross-program invocation and the System Program, checked against LiteSVM and Anza's interpreter. The VM and all 38 syscalls run in hand-set neural network weights. Here is the architecture, how each part works, how it was verified, how fast it is, and exactly what is not in the weights.

Topics: Svmllm, Solana, Sbpf, Agave, Litesvm, Neural Networks, Ai, Cryptography, Research, Verafy

A grid of 27 Verafy Improvement Proposals: VIP-0001 to VIP-0019 filled in as shipped, VIP-0020 to VIP-0027 outlined as upcoming
NOV 7, 2026

Verafy Harness v0.0.4: The Judiciary Signer, and Every VIP So Far

Verafy Harness v0.0.4-alpha adds the Judiciary Signer, a Squads multisig member that votes only on a sealed AI jury ruling bound to the exact transaction, along with policies and Covenant Launch, a harness registry, hardware attestation, model certification, bias profiles, an on-chain encoder, a validator commission covenant and asset transfer records. This post introduces the VIP system, lists every shipped and upcoming VIP, and covers the launcher, the local testnet, Testnet 0, the roadmap and the pilots.

Topics: Verafy Harness, Judiciary Signer, Squads, Solana, Vip, Covenant Launch, Ai Validators, Multisig, Verafy

Why now, in numbers: six developments listed beside three 1,024-token proof windows packed with small questions, and the line 120 proofs to 3
NOV 6, 2026

AI Oracles: Why Now, In Numbers

Why AI oracles are worth building now, argued with numbers: proof packing for Anthropic's FlashProve, the cost of an off-chain prover, fast System 1 screens like Jev, difficulty routing, auditable chain of thought, why a chain is the right record, and why three draft SIMDs mean it can all settle on Solana with no new chain.

Topics: Ai Oracles, Proof Packing, Flashprove, Provable Inference, Jev, Difficulty Routing, Chain Of Thought, Solana, Simd, Verafy Harness, Verafy

The verafyHarness logo beside a jury of five harnessed models, Claude, GPT, Gemini, Llama and Qwen, ruling at 94 percent against an 80 percent bar, with sealed blocks below
NOV 6, 2026

Verafy Harness: An AI Validator Anyone Can Run

Verafy Harness wraps one AI model in everything a validator does: intake, evidence checks, commit and reveal voting, a contract runtime, and a routing agent with a keyblock, a banker and an API router. It is the validator for Verafy Chain, a concept chain where contracts are written in words and there is no token. Five labs' models have sealed a block together.

Topics: Verafy Harness, Verafy Chain, Actually Smart Contracts, Ai Validators, Deepseek Harness, Verifiable Inference, No Token, Verafy

Four steps: a prompt program stored on chain, evaluators from many labs, a proof of inference, and a signed ruling the calling program verifies
NOV 5, 2026

Bringing Inference Programs To Solana

Three draft Solana Improvement Documents for inference programs: prompts stored on chain, executed by AI models off chain, with signed rulings, proofs of which model did the work, and challenges. All three drafts are embedded.

Topics: Actually Smart Contracts, Asc, Simd, Solana, Verifiable Inference, Provable Inference, Verafy

The ASC logo in ASCII, next to a rule and two rulings: a $250,000 withdrawal allowed 5 of 5, a $6 billion withdrawal refused 0 of 5
NOV 4, 2026

Actually Smart Contracts: rules in English, enforced by an AI jury

Actually Smart Contracts are rules written in plain English, inscribed on Solana, and enforced by a jury of AI models from different labs. Each vote is signed, and the contract acts only when enough jurors agree.

Topics: Actually Smart Contracts, Asc, Ai Jury, Smart Contracts, Solana, Iq Labs, Verafy

Title card on a dark background: Can you cram all of Solana into an LLM?
NOV 1, 2026

Can You Cram All of Solana Into an LLM?

We already built svmLLM, a tiny Solana-style VM inside neural network weights. So can all of Solana go in an LLM? We break down Agave versus the SVM, count the code, explain why a model needs a harness, and show why it would never keep up with consensus.

Topics: Solana, Agave, Svm, Sbpf, Neural Networks, Ai, Research

svmLLM title card: a tiny Solana VM inside neural network weights. 1,130,672 numbers, each written on purpose by a program, 0 trained.
NOV 1, 2026

svmLLM: A Tiny Solana VM Built Out of Neural Network Weights

svmLLM is a small Solana-style sBPF virtual machine written directly into neural network weights. Every one of its 1,130,672 numbers was set by code, not training. Here is how it works, how it was tested, and where it falls short.

Topics: Solana, Sbpf, Neural Networks, Ai, Pytorch, Research, Open Source

The Verafy Bias Detector home page
OCT 11, 2026

Introducing the Verafy Bias Detector (Beta)

The Verafy Bias Detector runs 565 versioned contentious questions against any model on OpenRouter, has a random cross-company LLM jury score the answers, and publishes a bias report card for each model.

Topics: Ai Bias, Llm Evaluation, Benchmarks, Openrouter, Solana, Verafy

The Verafy Inspector side panel open next to a Wikipedia article
OCT 11, 2026

Introducing the Verafy Inspector for Chrome

The Verafy Inspector is a Chrome side panel that finds the claims on any page, checks them against independent sources, and shows cliff notes and missing perspectives.

Topics: Fact-checking, Chrome Extension, Ai, Verafy, Media Literacy, Open Source

The Verafy Swarm Explorer
OCT 11, 2026

The Verafy Swarm Explorer: One Question, Many Models

The Verafy Swarm Explorer sends one question to many language models, uses other models as judges to check the answers, and shows where they agree and where they split.

Topics: Swarm Explorer, Fact-checking, Llm, Ai Bias, Verafy, Solana

Multiple application screens showing Verafy API integrations across different industries
AUG 29, 2025

44+ Powerful Use Cases for Verafy's AI Fact-Checking API

Explore comprehensive use cases for Verafy's fact-checking API across industries—from social media verification to enterprise compliance, Web3 platforms, and educational tools.

Topics: Use Cases, Api Applications, Fact-checking, Social Media, Enterprise, Web3, Education, Journalism, E-commerce

Developer coding with Verafy API and Solana integration
AUG 25, 2025

Verafy API, Getting Started: Trust-Powered Apps

Learn how to integrate Verafy's consensus-based fact-checking API with Solana blockchain to build transparent, trust-driven applications that combat misinformation.

Topics: Api, Developers, Solana, Blockchain, Ai, Fact-checking, Web3, Trust, Verification

Las Vegas Skyline with AI4 Conference Logo
JUL 13, 2025

Verafy CEO Rex St. John to Speak at AI4 Las Vegas: Advancing AI x Blockchain Innovation

Verafy CEO Rex St. John will be attending and speaking at AI4 Las Vegas this August, discussing Verafy's pioneering work at the frontier of AI x Blockchain integration.

Topics: Ai, Blockchain, Solana, Ai Governance, Ai X Blockchain, Crypto Infrastructure, Decentralization, Innovation

TruthNode joins DoubleZero validator testnet
JUL 9, 2025

TruthNode Joining DoubleZero Testnet: Advancing Solana Infrastructure with Verafy

Verafy’s validator, TruthNode, is joining the DoubleZero testnet, continuing our mission to secure the Solana network and pioneer sovereign infrastructure in the AI age.

Topics: Truthnode, Doublezero, Verafy, Solana Infrastructure, Validator Node, Solana Foundation, Jito Delegation, Builder Movement

Verafy team at Solana Accelerate
MAY 26, 2025

Verafy at Solana Accelerate: Unveiling the Future of Multi-AI Sovereignty

The Verafy team made our debut at Solana Accelerate, announcing our mission to build a sovereign multi-AI future and dropping our first video to outline the vision. Join the closed-beta and help us build the Internet of Truth.

Topics: Multi-ai, Ai Sovereignty, Verafy, Solana Accelerate, On-chain Truth, Ai Accountability, Builder Movement

Cypherpunk Truth Symbolism
MAY 13, 2025

Introducing Verafy Swarm Explorer Closed-Beta & Verafy Prime: Compare AI, Uncover Truth, and More!

Verafy Swarm Explorer enters closed-beta: compare up to 20 AIs, detect bias, and log results on-chain. Join Verafy Prime to access query credits and upcoming features.

Topics: Truth, Blockchain, Ai, Solana, Decentralization, Trustless Systems, Open Source, Ai Alignment