Basis report · 2026-06-09 · version 1
Base AI Claims Audit: What Verified, What Did Not
Integrity record
Verify: download the dataset, compute SHA-256 of the canonical JSON, and compare against the onchain record. Methodology →
Executive summary
The agent economy is moving fast. The numbers need a referee.
Basis reviewed the flagship public claims of nine Base-native AI projects — Bankr, Venice, Aeon, Gitlawb, PRXVT, Virtuals Protocol, HeyElsa, Moltbook, and Reppo — against onchain data, primary documents, and independent reporting. Each material claim was labeled verified, partially verified, unverified, disputed, or stale, with sources preserved.
The pattern is consistent: claims that are onchain-checkable or anchored in primary documents (token burns, raises, EIP authorship, acquisitions, repository activity) survive scrutiny. Claims about populations and volumes (agent counts, user counts, processed volume, GDP) almost never do — they are self-reported, sybil-sensitive, or contradicted by primary evidence.
This is not an accusation of bad faith. Counter-style metrics inflate by default when creation is free and verification is absent. The purpose of this audit is to make verification present.
Scope
Nine projects selected for prominence in the Base AI ecosystem as of June 2026. For each, Basis evaluated the project's most-circulated public claims: traction figures, token economics, technical milestones, partnerships, and standards involvement.
Evidence window: research conducted June 9, 2026, including a 104-agent adversarial verification pass (claims confirmed only on a 3-0 verifier vote) and targeted follow-up research for projects with no initially surviving claims.
Out of scope for this report: price prediction, investment recommendations, code audits, and any claim about intent.
Methodology
Claims were labeled verified (onchain, canonical primary document, or 2+ independent sources), partially verified (self-reported but plausible, single-source, or verified with a material caveat), unverified (no independent footprint found), disputed (primary evidence conflicts with the claim), or stale (dated evidence presented as current). Confidence levels describe the robustness of the evidence process, independent of the verdict. Only claims made by a project count toward its Claim Reliability Index; third-party findings are recorded as context.
The full scoring formula, sybil heuristics, and source-quality rules are published at basis.watch/methodology and versioned in the open repository.
Claim Reliability Index
| # | Project | Mix | Scored claims | Verified | Partial | Unverified | Disputed | Score | Grade |
|---|---|---|---|---|---|---|---|---|---|
| 01 | Venice | Claim mix: 3 verified, 2 partially verified, 0 stale, 0 unverified, 1 disputed — 6 scored claims. | 6 | 3 | 2 | 0 | 1 | 61incl. −6 disputed penalty | Grade C |
| 02 | HeyElsa | Claim mix: 1 verified, 1 partially verified, 0 stale, 1 unverified, 0 disputed — 3 scored claims. | 3 | 1 | 1 | 1 | 0 | 53 | Grade C |
| 03 | Bankr | Claim mix: 1 verified, 2 partially verified, 0 stale, 2 unverified, 0 disputed — 5 scored claims. | 5 | 1 | 2 | 2 | 0 | 44 | Grade D |
| 04 | Virtuals Protocol | Claim mix: 1 verified, 0 partially verified, 0 stale, 2 unverified, 0 disputed — 3 scored claims. | 3 | 1 | 0 | 2 | 0 | 41 | Grade D |
| 05 | Aeon | Claim mix: 0 verified, 2 partially verified, 0 stale, 2 unverified, 0 disputed — 4 scored claims. | 4 | 0 | 2 | 2 | 0 | 29 | Grade E |
| 06 | Reppo | Claim mix: 0 verified, 1 partially verified, 0 stale, 2 unverified, 0 disputed — 3 scored claims. | 3 | 0 | 1 | 2 | 0 | 25 | Grade E |
| 07 | Gitlawb | Claim mix: 0 verified, 2 partially verified, 0 stale, 1 unverified, 1 disputed — 4 scored claims. | 4 | 0 | 2 | 1 | 1 | 23incl. −6 disputed penalty | Grade E |
| 08 | PRXVT | Claim mix: 0 verified, 1 partially verified, 0 stale, 1 unverified, 1 disputed — 3 scored claims. | 3 | 0 | 1 | 1 | 1 | 15incl. −6 disputed penalty | Grade E |
| 09 | Moltbook | Claim mix: 0 verified, 0 partially verified, 0 stale, 0 unverified, 1 disputed — 1 scored claims. | 1 | 0 | 0 | 0 | 1 | 0†incl. −6 disputed penalty | Grade E |
† provisional — fewer than 3 scored claims
CRI measures claim reliability, not product quality, safety, or investment merit. Status labels describe evidence, not intent.
Key findings
- Onchain claims verify; population claims do not. Every claim that survived adversarial verification is either onchain-checkable or a primary-document fact. Every headline user/agent/volume figure examined was self-reported, contradicted, or unverifiable.
- Venice has the most credible token-economic loop reviewed: 33.75M VVV verifiably burned at the null address with revenue-funded buybacks — though the "42%" framing uses the most favorable denominator, and the circulating "3M+ users" figure runs ahead of Venice's own "2M+" page.
- Moltbook is the clearest documented case of population inflation: ~1.5M agent accounts mapped to ~17,000 humans (88:1) per Wiz Research — and Meta acquired the platform anyway, confirming that attention demand is real even when metrics are not.
- Gitlawb's live API returns numbers matching its claims, but the counters are project-operated, creation is free, and automated repo creation was directly observed. Its "agents earn onchain" claim is contradicted by its own pending-deployment status.
- PRXVT's flagship claim — leading ERC-8183 privacy sessions with the Ethereum Foundation — is contradicted by the EIP's author list and discussion record. The verifiable footprint is one merged community pull request. ERC-8183 itself is real (EF dAI + Virtuals authorship, verified).
- Virtuals' infrastructure claims verify (CCIP migration, ERC-8183 co-authorship); its "agent GDP" does not — the figure is self-reported, subsidized by up to $1M/month in incentives, and concentrated in a single agent serving ~3,700 users.
- x402 context: agentic transactions on Base exceeded 100M cumulative (Chainalysis), but independent analysis attributed roughly half of activity to incentive farming, and the value mix inverted away from micropayments — transaction counts alone are not adoption evidence.
- Self-reported growth can be internally consistent and still unverifiable: HeyElsa's volume trajectory ($20M → $503M+ in 12 months) is plausible and entirely first-party.
Project dossiers
Full dossiers — what each project does, its claim table, and our assessment — live on the project pages:
Full claim table
41 claims across 9 projects. Claims marked “context · not scored” were asserted by third parties or surfaced by Basis and do not affect any project's reliability score.
“Bankr is a Coinbase Ventures portfolio company (Base Ecosystem Fund).”
Coinbase's own blog lists Bankr with the asterisk denoting Coinbase Ventures portfolio status; the July 2025 Base Ecosystem Fund investment was independently reported. Investment amount undisclosed.
“Bankr generates meaningful fee revenue from token launches and trading.”
Third-party reporting referencing onchain fee data placed annualized revenue past ~$580K with ~$47.6K in 30-day fees as of February 2026. Public Dune dashboards exist but figures have not been independently audited by Basis. Revenue is real but modest relative to the token's ~$50M market capitalization.
Methodology note: Fee data is onchain-checkable in principle; Basis has not yet re-derived the annualized figure from raw settlement data.
“Thousands of agent token deployments via Bankr.”
No independent count of total Bankr token deployments was found in any public source. The Privy case study quantifies only a single token's reach ($DRB, 96k unique traders in under two weeks). One third-party analysis of ~25 Bankr-linked tokens found ~88% had under $10K cumulative volume with no recent activity, indicating most launches do not sustain a market.
Methodology note: Deployment counts are onchain-derivable from deployer contracts; a full re-derivation is queued for a future report.
“Bankr Agent (April 2026): in-browser agent workspace with portable skills and 410 MCP tools via gasless meta-transactions.”
The April 11, 2026 announcement and the public skills repository are confirmed. The specific tool count (410) and the gasless meta-transaction architecture are self-described and have not been independently exercised.
“Strategy libraries integrated into agent memory; agents becoming self-funding and trading nonstop.”
No public source using this phrasing or describing this feature was found. The closest documented feature set is the Bankr Agent skills/workspace announcement, which does not describe strategy libraries in agent memory.
“May 4, 2026: an attacker bypassed Bankr's safety block on Grok via a gifted Bankr Club membership NFT and obfuscated prompt injection, draining ~$150–200K in $DRB.”
Independently documented by SlowMist ("permission chain attack"), The Block, Giskard, and the OECD AI incident registry. Bankr stated ~80% of funds were returned. A separate March 2025 incident drained ~$330K. Recorded as context; this is a third-party finding, not a Bankr claim, and does not affect Bankr's Claim Reliability Index.
“33.75M VVV burned, with an ongoing revenue-funded buy-and-burn.”
The null address (0x000...000) holds 33,749,674.7 VVV onchain, matching Venice's dashboard. The March 2025 burn of ~32.6M unclaimed airdrop tokens was independently reported by The Block. Third-party tracker VeniceStats corroborates ongoing buybacks of roughly $100K/month.
“42%+ of VVV supply burned.”
The 42.05% figure is internally consistent but uses Venice's post-burn total supply (80.27M) as denominator. Against the 100M genesis supply the burn is ~33.75%; against Basescan's max supply field (~114M) it is ~29.6%. The burn is real; the percentage depends on a denominator choice that is the most favorable available framing.
Methodology note: Methodology-limited: percentage claims require a stated denominator to be comparable across projects.
“3M+ users.”
Venice's own marketing page stated "over 2 million users" as of June 2026, and its changelog announced the 2M milestone in late January 2026. Independent traffic data (Semrush: ~12.07M visits in April 2026, 74% direct) is consistent with an active base in the low millions but does not verify a 3M registered-user figure. The 3M+ claim, as circulated, runs ahead of Venice's own primary source.
Methodology note: Disputed because the claim conflicts with the claimant's own current primary source, not because usage is small.
“Major exchange listings including Upbit and Crypto.com.”
Upbit listed VVV with KRW/BTC/USDT pairs on May 12, 2026 (multiple independent outlets). Crypto.com App listing the same day. VVV also trades on Coinbase.
“Emissions cutting to 3M VVV/year.”
Venice announced a stepped schedule (May 1: 6M→5M; June 1: 5M→4M; July 1: 4M→3M), the fourth cut in under a year. The announcement is confirmed; execution of the final step postdates this check and is not yet verifiable.
“DIEM: each Diem provides $1/day of API credit, mintable only by locking staked VVV.”
Mechanism confirmed in Venice's documentation and observable onchain (DIEM ERC-20 on Base, ~37K supply). One caveat: the $1/day credit requires the DIEM itself to be staked (min 0.1), a condition the marketing shorthand omits.
“195 skills live across 8 categories.”
The official README (actively pushed as of June 9, 2026) lists 196 skills across 8 categories with a named catalog whose per-category sums check out. This is self-description from the project's own repository: the count of defined skills is supported; whether each skill functions as advertised has not been independently tested.
“Powering 40+ ecosystem projects.”
The main README lists approximately 16–18 named community skill packs and defers any broader count to a separate file. No independent enumeration supporting 40+ was found. In Basis's adversarial review, the claim that the README substantiates this figure was itself refuted.
“x402 pay-per-call settling in USDC on Base, MCP servers including base-mcp, and A2A support.”
All three integrations are documented verbatim in the README of an actively maintained repository. Documentation-level support is confirmed; settlement volume or live usage has not been independently measured.
“Self-healing across 8 categories (scored outputs, heartbeat → skill-health → skill-repair).”
The self-healing loop is described in the project's own documentation. No independent reliability testing, uptime data, or third-party reproduction was found.
“Aeon's README documents no native token; a community $AEON token launched via Bankr exists outside the repository.”
The README contains no token documentation. A Bankr-launched $AEON associated with the creator's account exists on Base. Recorded as context so that 'no token' is not over-inferred from the repository alone.
“10,757 repos and 33k+ agents on the network.”
Gitlawb's public node API returned agents: 32,888 / repos: 10,865 at check time, matching the claimed magnitudes. However, the API is operated by the project itself; repo and agent creation are free and unauthenticated in cost; and sampling the repo feed surfaced automated repositories (e.g. "cyber-213899", described as an automated repo, created seconds before the query). The counters are real but sybil-sensitive, and no independent de-duplication exists.
Methodology note: Counter-style metrics with zero creation cost are treated as sybil-sensitive by default.
“Live bounties; agents earn onchain.”
At check time the bounty escrow contract was not deployed to Base mainnet: the project's own bounty page lists the GitlawbBounty.sol address as "deploying soon" and the contracts README marks mainnet bounty/staking deployment "TBD — deploy pending." The 35 completed bounties shown are off-chain ledger entries with values between roughly $0.0001 and $0.09. "Agents earn onchain" is not supported by the project's own deployment status.
“OpenClaude MCP integration; Playground apps publish directly to nodes.”
The node software ships an MCP server (25 tools) and OpenClaude lists Gitlawb as a sponsor and integration. OpenClaude is maintained by the same organization, so this is an internal integration rather than third-party adoption. OpenClaude itself shows 28.5K stars but a 0.77% watcher ratio, an unusually high fork/star ratio, and a README acknowledging it originated from the Claude Code codebase — which is not open source — creating unresolved provenance and license questions (README badge: MIT; GitHub API: NOASSERTION).
“Token-powered network economics: staking, storage fees, bounty fees, node rewards.”
$GITLAWB exists on Base (verified onchain; ~$9M market cap at check, −79% from May 2026 high; 7,088 holders). The economic mechanisms described — staking tiers, storage fees, the 5% bounty fee, fee distribution — are deployed on Base Sepolia testnet only. On mainnet, the token currently has no active protocol utility.
“ERC-8183 ("Agentic Commerce") is a real Draft ERC authored by the Ethereum Foundation dAI team and Virtuals Protocol.”
ERC-8183 exists on the canonical EIP repository: Draft, created February 25, 2026, authored by Davide Crapis (Ethereum Foundation dAI) with three Virtuals Protocol co-authors. It defines a job-escrow primitive with extensible hooks. Recorded as context for evaluating PRXVT's involvement claim.
“Leading ERC-8183 privacy sessions with the Ethereum Foundation and Virtuals.”
PRXVT is not an author of ERC-8183, appears zero times in the canonical Ethereum Magicians discussion thread (12 participants), and no Ethereum Foundation source mentions PRXVT. The verifiable footprint is narrower: one PrivacyHook pull request merged on May 11, 2026 into the community hooks repository linked from the EIP, and an ecosystem-partners listing merged into the x402 repository in January 2026. A merged community PR is meaningful but does not support "leading privacy sessions with the Ethereum Foundation." The "Virtuals" relationship is, in verifiable part, a token launch on the Virtuals launchpad.
“px402: unlinkable x402 payments via ZK proofs and ERC-4337 burner accounts.”
The TypeScript SDK exists and describes the claimed architecture (Groth16, fixed-denomination notes, 4337 burners). However, the repository has not been pushed since November 25, 2025; everything demonstrated runs on Base Sepolia testnet; and no security audit of the ZK scheme was found. npm shows ~37K downloads, a metric that is trivially inflatable and sits oddly against 22 GitHub stars.
“Building the blackbox standard for agent commerce.”
No standards-track document, working group, specification draft, or independent reference connecting "blackbox standard" to PRXVT or any standards body was found. The phrase has no footprint outside the project's own marketing.
“Multi-chain expansion through Chainlink CCIP.”
Confirmed: as of June 4, 2026, Virtuals is migrating $700M+ in VIRTUAL cross-chain transfers from LayerZero to Chainlink CCIP as exclusive cross-chain infrastructure, corroborated by multiple independent outlets. Context: the $700M figure is token value secured, not usage or revenue; the announcement contains no agent, transaction, user, or revenue metrics; and the migration was announced/in-progress, not complete, at check time.
“Agent economy expanding past multi-million GDP.”
Reported aGDP figures ($470–479M cumulative, 1.77M completed jobs, 18,000+ agents) trace to project materials and exchange content; no independent measurement exists. Three caveats are documented: the figures are self-reported; the Revenue Network program distributes up to $1M/month in incentives to transacting agents, so organic versus subsidized volume is not separable from public data; and activity is concentrated — the top agent claims roughly half of all aGDP from ~3,700 unique users. Notably, Virtuals' own June 2026 CCIP announcement contains no GDP figure.
Virtuals Protocol · #virtuals-revenue-trajectory“Virtuals launchpad revenue declined ~96% from its peak (daily revenue ~$1.02M → ~$34.8K; new agent creation 1,300/day → under 10/day).”
Independent reporting by Decrypt with onchain-derived figures. Cumulative protocol revenue exceeds $60M historically. Recorded as context: launchpad revenue is real and large historically, and sharply reduced from peak.
“Weekly LLM trading competitions.”
Referenced in project communications; no independent coverage, results data, or onchain competition registry was located during this check.
Virtuals Protocol · #virtuals-erc8183-authorship“Virtuals Protocol co-authored ERC-8183 with the Ethereum Foundation dAI team.”
Three Virtuals Protocol authors appear on the canonical EIP alongside Davide Crapis (EF dAI). Recorded as positive context: standards authorship is among Virtuals' strongest verifiable assets.
“$3M raised from M31 Capital and Coinbase Ventures.”
Announced June 5, 2025: $3M combined pre-seed/seed led by M31 Capital with Coinbase Ventures' Base Ecosystem Fund, MH Ventures, Absoluta Cap, 2Shares, and Levitate Labs. Coinbase's own blog marks HeyElsa as a CV portfolio company.
“$500M+ volume processed.”
Every volume figure in the public record originates from HeyElsa's own materials: $20M at the June 2025 raise, ~$168M reported December 2025, $503M+ displayed on heyelsa.ai as of June 2026. The trajectory is internally consistent and the implied ~25x growth in 12 months is possible for an execution product, but no independent onchain audit or third-party measurement of routed volume was found. Intermediate milestones in third-party coverage trace back to HeyElsa self-reporting.
Methodology note: Routed-volume claims are onchain-derivable if router contracts are disclosed; HeyElsa has not published a contract-level breakdown.
“18M+ prompts; 945K+ wallets.”
Prompt and wallet counts appear only in HeyElsa's own marketing surfaces. Prompts are an off-chain metric with no public verification path; the wallet figure has no published methodology (created vs. connected vs. active).
“Millions of agent accounts; the front page of the agent internet.”
Wiz Research found the reported ~1.5M agents corresponded to roughly 17,000 human accounts — about 88 agents per person — with no rate limiting and no verification that an "agent" was AI-operated; anyone could register large numbers programmatically. Because exposed credentials allowed humans to post as agents, viral "agent culture" screenshots are partly attributable to human operation. Account records existed; the implied population of millions of autonomous agents is contradicted by the primary investigation.
“February 2026: misconfigured database exposed ~4.75M records including 1.5M agent API tokens, ~35,000 emails, and private messages.”
Wiz Research documented a Supabase API key hardcoded in client-side JavaScript with no Row Level Security, granting unauthenticated read/write to production data. Remediated within hours of disclosure. Corroborated by Fortune, Infosecurity Magazine, and others.
“Meta acquired Moltbook (March 10, 2026); price undisclosed.”
Confirmed by Axios (exclusive), TechCrunch, CNBC, and CNN. Founders Matt Schlicht and Ben Parr joined Meta Superintelligence Labs. No outlet disclosed a price; any circulating figure is unsupported. The site remained operational in a transitional state at check time.
Moltbook · #moltbook-base-relevance“Moltbook is a web2 platform with no official token; the $MOLT token on Base is unofficial.”
The platform ran on a conventional web2 stack (Supabase + OpenClaw agents). No acquisition coverage mentions a token or Base deployment by the team. A $MOLT memecoin exists on Base with no official ties — relevant for anyone treating $MOLT as exposure to Moltbook.
“$20M strategic capital commitment from Bolts Capital (April 2026).”
The announcement exists and was syndicated (Chainwire, CoinDesk press-release section). However: it is a press release, not independent reporting; the structure of the "commitment" is undisclosed (committed capital is not deployed capital); and Bolts Capital is a small fund with no top-tier track record. The earlier $2.2M seed (Protocol Labs, CMS, CV VC) follows the same press-release-only pattern. The announcement is confirmed; the economic substance is not.
“Onchain evaluation and datanets via Orquestra.”
"Orquestra" appears only in aggregator/marketing-derived copy. It is not mentioned on reppo.xyz, docs.reppo.ai, the project GitBook, the whitepaper link, or the $20M press release, and no repository by that name exists under Reppo-Labs. The datanet design itself is documented; the "Orquestra" framework has no primary-source footprint.
“Over $2M trading volume in the past month.”
Stated in the project's funding press release. No independent dashboard, onchain derivation, or third-party measurement was found. Token market data (REPPO on Base via the Virtuals launchpad, ~$5.3M market cap at check) is verifiable, but datanet trading volume is not.
“Reppo's open-source development is real and active.”
The Reppo-Labs GitHub organization (created July 2024) shows continuous activity, with reppo-cli pushed on the day of this check. Documentation describes the datanet mechanism (publisher/voter roles, veREPPO, 48-hour epochs, minting on Base). Recorded as positive context: the engineering footprint is genuine even where traction claims are not yet verifiable.
Limitations
- Basis labels claims, not projects. A disputed claim does not make a project a failure, and a verified claim does not make it an investment.
- Unverified means no independent footprint was found as of the check date — not that a claim is false. Projects are invited to submit evidence via corrections@basis.watch; the correction process is public.
- Sybil heuristics are estimates with stated assumptions, published in full in the methodology. Raw figures are always reported alongside adjusted ones.
- Several x402 datasets in circulation (facilitator marketing, Chainalysis, independent analysts) are mutually inconsistent; where they conflict, this report preserves the conflict rather than selecting a side.
- Web traffic estimates, npm download counts, and social metrics are weak proxies and are labeled as such wherever used.
Citations
- Basescan — VVV token, null-address balancehttps://basescan.org/token/0xacfE6019Ed1A7Dc6f7B508C02d1b04ec88cC21bf
- The Block — Venice airdrop burnhttps://www.theblock.co/post/345912
- Wiz Research — Moltbook database exposurehttps://www.wiz.io/blog/exposed-moltbook-database-reveals-millions-of-api-keys
- Fortune — Moltbook security reportinghttps://fortune.com/2026/02/03/moltbook-ai-social-network-security-researchers-agent-internet/
- Axios — Meta acquires Moltbookhttps://www.axios.com/2026/03/10/meta-facebook-moltbook-agent-social-network
- EIP-8183 — Agentic Commerce (Draft)https://eips.ethereum.org/EIPS/eip-8183
- Ethereum Magicians — ERC-8183 discussionhttps://ethereum-magicians.org/t/erc-8183-agentic-commerce/27902
- EIP-8004 — Trustless Agents (Draft)https://eips.ethereum.org/EIPS/eip-8004
- PR Newswire — Virtuals CCIP migrationhttps://www.prnewswire.com/news-releases/virtuals-protocol-migrates-700m-virtual-token-from-layerzero-to-chainlink-ccip-to-enable-secure-cross-chain-payments-for-ai-agents-302790918.html
- Decrypt — Virtuals revenue analysishttps://decrypt.co/309495/virtuals-protocol-revenue-crashes-as-ai-agent-demand-sinks
- HeyElsa — $3M raise announcementhttps://blog.heyelsa.ai/heyelsa-raises-3m-to-build-ai-stack-for-crypto/
- Coinbase — Onchain AI agents blog (CV portfolio markers)https://www.coinbase.com/blog/the-rise-of-onchain-ai-agents-apps-and-commerce
- SlowMist — Grok permission-chain analysishttps://slowmist.medium.com/behind-the-grok-exploitation-an-analysis-of-ai-agent-permission-chain-abuse-4d832d1bfc73
- The Block — Bankr incident coveragehttps://www.theblock.co/post/346027/
- Chainalysis — x402 agentic payments adoptionhttps://www.chainalysis.com/blog/x402-agentic-payments-adoption/
- CoinDesk — x402 demand analysishttps://www.coindesk.com/markets/2026/03/11/coinbase-backed-ai-payments-protocol-wants-to-fix-micropayment-but-demand-is-just-not-there-yet
- Coinbase Developer Platform — x402 docshttps://docs.cdp.coinbase.com/x402/welcome
- Gitlawb node API — live statshttps://node.gitlawb.com/api/v1/stats
- Gitlawb contracts — deployment statushttps://github.com/Gitlawb/contracts
- Aeon — repository and skill cataloghttps://github.com/aaronjmars/aeon
- Venice changelog — 2M user milestonehttps://featurebase.venice.ai/changelog/veniceai-change-log-january-27-2026-february-3-2026
- Crypto Briefing — Upbit VVV listinghttps://cryptobriefing.com/upbit-venice-token-vvv-listing/
- Chainwire — Reppo capital commitment press releasehttps://chainwire.org/2026/04/23/reppo-foundation-secures-20m-capital-commitment-to-solve-training-data-bottleneck-using-prediction-markets/
- Reppo docshttps://docs.reppo.ai/
- Privy — Bankrbot case studyhttps://www.privy.io/blog/bankrbot-case-study
- Dune — Bankr fee dashboardshttps://dune.com/bankr
Legal
This report is published for informational purposes only. It is not financial, investment, legal, or tax advice, and it is not a recommendation to buy or sell any asset. Statements about claims reflect the state of cited public evidence as of the check date and may change as new evidence emerges. Status labels describe evidence, not intent. Projects named in this report may submit corrections with supporting evidence to corrections@basis.watch; substantiated corrections are applied and logged in a public changelog.