A growing record · independent · unsolicited
We take a chance at optimizing how real institutions work.
Nobody asked us to. Each entry is an independent, passive, evidence-based look at a real organization's public surface, what a person hits trying to get something done, and what an AI sees trying to represent them, and the fix we'd bring. Same root cause both ways: systems built once, for a narrow purpose, never made legible to anyone. We do the work on our own initiative, honestly, and keep it.
The portfolio: three tracks
Segmented, not ranked. Every engagement is kept and elevated; the split just names what each one is.
Institutional EWP-integration concepts
Independent concept demonstrations of where the Ecological Work Protocol, a funded, proof-verified, on-chain-settled work-packet layer, could plug into a large institution's public surface. Each is unsolicited and unaffiliated.
OpenAI
Publicly forecasts labor displacement but has no funded, proof-verified destination wired to route searching workers into real work, EcoWealth's Work Protocol proposed as that on-ramp. Partnership concept, 8 cited positions.
EWP-integration concept Open the review →Anthropic
The Economic Index maps where work goes; the map still needs a funded, proof-verified destination. EcoWealth proposed as that destination, reachable over Anthropic's own MCP standard. Partnership concept, built on Claude.
EWP-integration concept Open the review →Chase
Four consumer vendor stacks with zero machine-readable product data and a PDF-bound $2.5T climate commitment, yet JPMorgan Payments already ships a live llms.txt and Kinexys runs on Base. Extend that pattern + add a proof-of-outcome layer. 7-finding deep audit.
EWP-integration concept Open the review →U.S. Bank
Clean build and a genuinely ahead payments arm, Elavon already ships a real production llms.txt (286 lines, text/markdown), yet the consumer bank renders 0 machine-readable product/ESG data, ships no consumer llms.txt (404), and its $50B environmental commitment is PDF-bound. U.S. Bank already custodies on-chain stablecoin reserves; extend Elavon's pattern to consumer + ESG and add a proof-of-outcome layer on Base. 4-finding brief, with a self-correction (Elavon SPA soft-200s aren't real files).
EWP-integration concept Open the review →IRS
Independent Evidence Workpaper concept on fabricated data only, casework evidence is still mail/upload-shaped, status is a phone call, and there is no case object an AI can verify. NOT an IRS project; no adoption claimed. 7-finding deep audit.
concept · not affiliated Open the review →Inter-Con Security
A physical-security firm whose own email is spoofable (DMARC p=none, no security.txt) and whose proof-of-service is self-attested, verifiable proof is exactly their business. Proposed as a dedicated provisioner. 6 findings, DNS-evidenced.
EWP-integration concept Open the review →Skanska
Strong SBTi-accredited ESG substance locked in PDFs, a subcontractor front door returning 502 / VPN-required, and greenfield agent-discoverability. EWP + Agent Kit plug into bids, ESG data, and subcontractor coordination. 4 findings.
EWP-integration concept Open the review →Turner Construction
Industry-leading safety culture (Building L.I.F.E., ISO 45001, the free SafeT Coach advice layer) with no machine-verifiable proof that work was done safely beneath it, self-attested subcontractor prequalification, and zero agent-readable structured data (llms.txt a true 404). EWP proposed as the hazard-gated, proof-verified, on-chain safety-settlement layer. 4 findings.
EWP-integration concept Open the review →Kaiser Permanente
Built the country's most complete social-health referral network, Thrive Local, the Community Support Hub (~350k connected), a $50M Food Is Medicine commitment, and is the first carbon-neutral U.S. health system (2020). But nothing proves the help arrived: no agent front door on any host (kp.org/llms.txt 302, about./llms.txt 301 to an error page, healthy./llms.txt 410, agent-card 404, control-verified), commitments carry only 1 ld+json block (a VideoObject, not data), and kp.org DMARC is monitor-only (p=none/sp=none) while retail peers run p=reject. EWP proposed strictly non-medical as the proof-of-outcome layer under Thrive, verified need, verified done, verified funded; clinical questions always route to a clinician. 4-finding brief.
concept · not affiliated Open the review →Walmart
Bet on agentic commerce with Sparky (its own AI shopping assistant, June 2025) and finished Project Gigaton six years early (~1.19 Gt), yet ships no open agent surface (walmart.com/llms.txt 404, control-verified; agent-card 404; developer domain 404/401) while direct competitor target.com/llms.txt returns 200, and the corporate sustainability page carries 0 ld+json (the 50M-acre regenerative pledge is prose, self-reported, unverifiable per place). Credit where due: DMARC p=reject and a real security.txt. With ~4,600 US stores (~90% of Americans within 10 miles), every store is a PLACE, EWP proposed as the open agent door + the per-store restore proof layer under the regenerative commitment. 4-finding brief.
concept · not affiliated Open the review →Target
The only institution in this portfolio that already ships a real agent front door, target.com/llms.txt returns 200, a genuine agent map of category/browse/product URL grammar. But it stops at shopping: 0 matches for 'Target Forward'/sustainability, self-labeled 'Phase 1', no agent-card (404), no ai.txt (404), no verifiable-outcome surface; and the Target Forward strategy page carries only 1 ld+json (a WebPage, not the goals, graded in prose achieved/progressing/evolving). Credit: DMARC p=reject, HSTS includeSubDomains. The easiest first buyer here, the work is the next commit on a file Target already maintains: extend the llms.txt with Target Forward + an agent-card, and add EWP as the per-store proof-of-outcome layer (regenerative = restore work, inclusive = the step-1 ladder). 4-finding brief.
concept · not affiliated Open the review →The company that authored the agent-discovery standard, A2A, launched April 2025 and donated to the Linux Foundation with AWS/Microsoft/Salesforce/SAP/ServiceNow, is not discoverable by it: google.com, blog.google, sustainability.google, gemini.google.com and developers.google.com all 404 for /llms.txt and /.well-known/agent-card.json (soft-404 controls confirm real 404s), while a2a-protocol.org/llms.txt (Google's own standard) returns 200. Credit-first: Google's security.txt is best-in-class (VRP, expiry 2030), so the lift is small and in-character. The 24/7-carbon-free-energy-by-2030 goal + 2026 Environmental Report are prose/PDF, not a queryable per-project outcome object. EWP proposed as a live A2A/x402 reference implementation AND the funded, proof-verified destination a Gemini agent can route ecological work to. 6-finding evidence-led brief, ask leads with a $0.10 provable call.
concept · not affiliated Open the review →YouTube
The largest creator-labor market on earth, $100B+ paid to creators since 2021, 2M+ in the Partner Program, and (by YouTube's commissioned Oxford Economics study) 490,000 US full-time-equivalent jobs and $55B in GDP in 2024, runs with no proof-of-outcome layer beneath the payments: monetization proves views and watch-time, not that a creator's filmed cleanup/planting/build was real. And there's no agent surface (youtube.com/llms.txt + /ai.txt + /.well-known/agent-card.json all 404; soft-404 control confirms). Credit-first: youtube.com ships a valid security.txt and clean robots.txt. EWP proposed as the funded, proof-verified, on-chain-settled outcome layer for creator labor, with the three ladders (Build/Grow&Cook/Restore) as filmable, serialized content, each episode settling a real packet. 5-finding evidence-led brief; ask leads with a $0.10 provable call.
concept · not affiliated Open the review →Apple
One of the most credible corporate climate programs on earth, Apple 2030 (carbon neutral, -75% vs 2015), the Restore Fund ($200M in 2021, +$200M in 2023, expanded 2025), 38M+ MWh of supplier clean energy in 2025, publishes its proof as an annual PDF (the 2026 Environmental Progress Report), not as per-project machine-verifiable objects. And apple.com exposes no agent surface (llms.txt / ai.txt / agent-card.json all 404; soft-404 control confirms) even though Apple already maintains other machine-readable well-knowns (security.txt with Apple Security Bounty + expiry 2027, apple-app-site-association, sitemaps), so the lift is small and in-character. EWP proposed as the per-project, on-chain proof-of-outcome layer beneath the Restore Fund, plus an agent-readable pattern for product/trade-in/repair data (France's AGEC per-product sheet is the on-ramp). 5-finding evidence-led brief; ask leads with a $0.10 provable call.
concept · not affiliated Open the review →Amazon
Amazon already ships the two hardest pieces, a production llms.txt on developer.amazon.com (200, 950 lines, 69 KB) and the most advanced agentic-commerce surface in retail (Rufus / 'Buy for Me'), yet every consumer, sustainability, and Climate Pledge llms.txt is a real 404, the storefront exposes no agent card, the sustainability page renders 0 structured data, and the 656-signatory Climate Pledge is a PDF number with no verified outcome under it while Amazon's own emissions rose 16%. Extend the developer pattern to the consumer + ESG surfaces and add a proof-of-outcome layer on Base, place-native to 350 fulfillment centers. 5-finding brief, evidence reproducible.
concept · not affiliated Open the review →Meta
The best security posture in the portfolio (CSP nonces, X-Frame DENY, DMARC reject) wrapped around a striking gap: the open-weight-AI company's only advertised llms.txt (llama.com/llms.txt → 301 → developer.meta.com/ai/llms.txt/) resolves to an HTTP 400 error page, and no agent-readable llms.txt resolves on any Meta surface. Facebook Groups, the largest community-mobilization surface on earth, has no proof-verified work layer behind 'community action,' Llama agents have brains everywhere but no on-chain work destination, and net-zero/water-positive 2030 are real but PDF-bound. Serve a real file where the redirect already points, expose a priced agent-card any Llama agent can call over x402, and add a proof-of-outcome layer under the water number. 5-finding brief, evidence reproducible.
concept · not affiliated Open the review →X
A thoughtful static-file estate (RFC 9309-aware robots.txt, Cloudflare bot management, DMARC reject) wrapped around a soft-200 blind spot: every agent-shaped path, llms.txt, ai.txt, .well-known/agent-card.json, and a random control, returns the identical 265,895-byte React app shell, so a machine sizing up X gets a blank page, not a file. X Money is live (Visa Direct, 6% APY, instant creator payouts) with no machine-payable destination to settle real work against, X's real-time coordination has no proof layer, and the PGP-signed security.txt is expired (2024-01-01) and still Twitter-branded. Replace the shell with real files, expose a priced agent-card any X-Money-native / Grok agent can call over x402, and add a proof-of-outcome layer on Base. 5-finding brief, evidence reproducible.
concept · not affiliated Open the review →GitHub
The world's largest work-coordination platform (~150M developers) already runs the full loop for code, Issue → claim → PR → review → merge, and now layers verifiable proof-of-outcome (artifact attestations: Sigstore + Rekor transparency log, SLSA provenance) and a $50M+ payment rail (Sponsors) on top, with a coding agent that turns an Issue into a PR autonomously. It ships a real llms.txt (28 KB, both hosts) and security.txt, but no A2A agent-card (true 404 vs control). EWP is that exact lifecycle for PHYSICAL ecological work, settled on-chain, an agent that already ships your code can fund and verify it keyless over x402. 5-finding brief, hardest HTTP evidence first, credited baseline.
concept · not affiliated Open the review →NVIDIA
The world's most valuable company (~$4.9T) is also one of the most agent-ready surfaces on the web, a maintained llms.txt hierarchy (www index → docs 'updated 10 July 2026' → developer 28 KB → locale), a real A2A agent card on developer.nvidia.com, and a robots.txt that explicitly Allows 50+ named AI crawlers into a /*.llms.txt$ + /*.md$ lane. But the buildout's #1 question, its energy/water footprint, is PDF-bound and self-attested (FY25 Scope 3 up 3.5M→6.9M MT CO2e), the corporate host's entire /.well-known/ tree is 403, and the developer markdown skill only partly resolves. EWP + Regen Network add the missing layer: a verified, on-chain ecological-retirement receipt per unit of compute. Credit-first, 5-finding brief, hardest HTTP evidence first.
concept · not affiliated Open the review →Costco
The most trusted labor model in retail (~341,000 employees, ~93% one-year retention, top clerks near $31.90/hr) with real Kirkland Signature commitments (100% RSPO palm oil, zero-net-deforestation), but the commitments are PDF-bound and self-attested, and the site is actively invisible to AI agents: a default-UA request returns 000 (blocked by Akamai Bot Manager before robots.txt loads), and /.well-known/agent-card.json returns HTTP 500 (not even a clean 404). A brand whose whole moat is trust is the natural carrier of proof-of-outcome receipts, on Kirkland supply-chain commitments and place-based restore work around every warehouse, funded and settled on Base. 4-finding brief, hardest HTTP evidence first, credited baseline.
concept · not affiliated Open the review →The Home Depot
World-class retail engineering, but to a non-browser AI agent, homedepot.com is a 403 'Access Denied' wall (AkamaiGHost) on home, product, and category alike, while robots.txt invites crawlers; the Path to Pro labor surface is reachable yet machine-illegible (corporate home ships 0 JSON-LD, pathtopro.com is a soft-404 SPA), and no llms.txt / agent-card exists on any Home Depot property. Path to Pro trains people for a ~400k-job trades gap but has no proof-of-work layer connecting a trained worker to a verified, paid job. EWP proposed as that Build-ladder proof layer, funded, hazard-gated, proof-verified, settled on Base. 4-finding brief; palette from THD's documented brand orange (live CSS wall-blocked, stated openly).
concept · not affiliated Open the review →Bechtel
A top-2 U.S. contractor and the world's largest privately-held EPC firm with a genuine safety record, verbatim 'zero lost time incidents and zero recordable incidents' goal, stop-work authority, 2023 LTIR 0.014 / RIR 0.163, yet no way for a machine to verify a single safe task: bechtel.com/llms.txt + ai.txt + every .well-known agent path are true 404s (soft-404 control confirmed), the homepage and the /approach/ safety+ESG hub each carry exactly one JSON-LD block (generic Yoast graph, no Organization/Project/Dataset), and ESG lives entirely in prose (100% renewable by 2025 at Quebrada Blanca; safety x45 on /approach/, 0 in structured data). EWP proposed as the hazard-gated, proof-verified, on-chain safety-settlement layer under the zero-incidents goal. 4-finding safety-first brief; palette from Bechtel's live theme CSS.
concept · not affiliated Open the review →Kiewit
The most agent-forward of the giants, alone among peers Kiewit ships a LIVE llms.txt (HTTP 200, 2,495 bytes, Yoast v26.9) and an Organization JSON-LD node, yet the door leads to an empty room: the llms.txt is a passive SEO content index with no capabilities/actions, /.well-known/agent-card.json + /ai.txt are 404, the /careers/ front door to a ~34,500-person self-perform craft workforce carries 0 JobPosting schema, and the 'Nobody Gets Hurt' safety record is prose only. Because Kiewit self-performs and is employee-owned, proof-of-safe-completion is an internal quality asset (not just a subcontractor gate) and pay-follows-proof is native. EWP proposed as the verification room behind the open door, hazard-gated craft packets + Restore-ladder ecological packets (their own Salton Sea / Cleaning-the-Creek work), settled on Base. 4-finding brief; palette = Kiewit black + signature yellow, Industry typeface.
concept · not affiliated Open the review →Uber
The largest 'bounded task, proven at a place, paid' machine in the US, ~1M active US drivers/couriers running the exact shape of an EWP work packet, yet its public front door ships no static agent-legible layer (llms.txt 404; the Cloudflare edge 406-blocks every /.well-known/ path despite robots.txt 'allow: /.well-known/'), and its real 2040 zero-emissions pledge has no machine-verifiable proof-of-outcome object. EWP proposed as the verifiable proof + Restore-packet on-ramp: restore work claimable between rides in the same city geography, settled on-chain. 4-finding brief, DNS + HTTP-status evidenced.
concept · not affiliated Open the review →ABM Industries
One of the largest US facility-services labor employers (~113,000 employees, $8.75B FY2025, NYSE: ABM) and genuinely ahead of its sector on machine-legibility, it already ships a real /llms.txt (200, 146KB) and enforces email auth (DMARC p=reject, SPF -all). But that llms.txt is a flat auto-generated link dump (0 capability sections) and every service page's JSON-LD is generic Organization/WebSite only, no Service/Offer schema, so an agent can't tell landscaping from microgrids; proof-of-service across 113k workers is self-attested timesheets (the Inter-Con pattern); and sustainability commitments are prose with no proof-of-outcome layer. EWP proposed as the tamper-evident service-receipt + restore-ladder layer above the discovery file ABM already publishes. 4-finding brief, HTTP + DNS + rendered-schema evidenced.
concept · not affiliated Open the review →UPS
The company that made machine-verifiable chain-of-custody a global standard, the signature, Complete View proof-of-delivery, applies that rigor to everyone's packages but not to its own commitments: its carbon-neutral-2050 pledge, its 12,000+ vehicle Rolling Lab, and its contractual promises to ~490,000 workers ship without a proof-of-outcome layer, and its public web estate refuses AI agents entirely (www.ups.com resets non-browser connections, curl 000 on every path incl. robots.txt; about.ups.com 403s). Every route is a place-sequence, so the delivery-proof loop IS the EWP work-packet loop. EWP proposed as the signature turned inward: proof-verified, on-chain outcome receipts a regulator or union rep replays independently. 4-finding brief, connection-status + DNS evidenced.
concept · not affiliated Open the review →Robinhood
Robinhood Chain launched (2026-07-01) as a self-described "permissionless, AI-native Layer 2 built for real-world assets" where "AI agents can trade, swap, lend, and transact", yet robinhood.com serves those agents a soft-200 (agent-card.json / ai-plugin.json / x402 all return an identical 8,084-byte HTML shell, not JSON), ships no llms.txt (confirmed absent vs an S3 control), and its security.txt expired two days before capture. Tokenized stocks exist; proof-verified real-world WORK, a self-documenting RWA that writes its own on-chain provenance, is the asset class nobody has tokenized. EWP proposed as that primitive, already live on the same L2 family. 6 findings, curl-reproducible.
concept · not affiliated Open the review →Coinbase
The strongest agent surface in this portfolio, credited first: the company that authored x402 actually ships a real consumer llms.txt (www.coinbase.com/llms.txt, 200/5,098 B) and a real A2A agent card on docs (docs.cdp.coinbase.com/.well-known/agent-card.json, application/json, protocolVersion 0.3). The refinements are precise, the consumer security.txt expired ~2 years ago (Expires 2024-06-01), and www.coinbase.com ships no A2A card and no /.well-known/x402 (true 404s) despite authoring the standard. The genuinely new thing: every indexed x402 merchant sells a digital API, none settle a verified PHYSICAL outcome. EWP is that proof-of-outcome merchant, standard-conformant and self-settled (no facilitator), live on Base. 5 findings, curl-reproducible.
concept · not affiliated Open the review →Base
Base's 2026 mission is 'a million builders, human or agent', and docs.base.org lives up to it: a real 61 KB llms.txt naming 'AI Agents', an llms-full.txt, and a real A2A agent card (Base MCP, x402). Credited first. But base.org, the flagship builder-showcase, is a true-404 desert: llms.txt / robots.txt / agent-card.json / ai.txt all 404 (control-confirmed), so an agent can't discover a single builder there; the mission words 'agent / builder / million / onchain' sit in prose with no machine-readable twin. EcoWealth runs a complete real-world-work settlement loop LIVE on Base mainnet (contract 0x76c17C…A14B, source-verified, workId 14 settled), exactly the 'human or agent' builder the mission wants, and undiscoverable via base.org today. The review is a worked example + the one surface base.org must inherit from its docs. 4 findings, curl-reproducible.
concept · not affiliated Open the review →Southwest Gas
A PHMSA-regulated gas utility (2.28M customers, 34,000+ mi across AZ/NV/CA) whose daily work, leak surveys, COYL survey-and-replace, $408M/yr integrity replacement, already has the exact shape of a proof-verified work packet, yet the mandatory completion records stay trapped as trust-me PDFs: homepage + sustainability page carry 0 ld+json, every agent well-known is a true 404, the sitemap ships 7,049 all-relative URLs (violates the protocol), and the root has no CSP/XFO/security.txt. EWP proposed as the field-work verification layer, hazard-gated authoring, photo+GPS+signature proof, on-chain settlement receipt, the regulatory records they must keep anyway, produced automatically. Flagship of the 4-target energy batch; 5-finding brief. Teach-first ask.
concept · not affiliated Open the review →NextEra Energy
The world's largest wind + solar generator (~73 GW, 30 GW backlog, 81 GW target by 2027, ~$184B market cap) builds every gigawatt with contractor crews (Build ladder) on land under easement (Restore ladder), thousands of proof-of-completion events with no per-project machine-verifiable proof layer. Its security baseline is genuinely strong (HSTS preload, XFO, CSP frame-ancestors), but the whole /.well-known/ agent tree returns a 403 wall (agents fail closed, not merely 404), llms.txt is a true 404, the homepage carries 0 ld+json, and there is no security.txt. EWP proposed as the per-project proof layer, hazard-gated safety-completion + easement restoration packets, settled on-chain, the use-of-proceeds argument settled at the project level. 4-finding brief. Teach-first ask citing NextEra's own AI-load-boom thesis.
concept · not affiliated Open the review →Edison International / SCE
Southern California Edison spends $6.2B on its 2026-2028 Energy-Safety-approved Wildfire Mitigation Plan (LiDAR + satellite vegetation management, covered conductor, inspections, PSPS), yet proof that a specific clearance or inspection was actually completed at a specific span and date, literally the post-ignition liability question, has no public verifiable object. edison.com deliberately welcomes AI crawlers in robots.txt (GPTBot, ClaudeBot, PerplexityBot, Google-Extended...) but then serves them nothing: llms.txt is a true 404, no agent card, 0 ld+json on both edison.com and sce.com; edison.com's HSTS is a 5-minute window and neither domain has a security.txt. EWP proposed as the wildfire-completion proof layer, hazard-gated clearance/inspection packets, GPS+photo+signature, settled on-chain, the CPUC/Energy-Safety records they must keep anyway, made court-grade. LA home turf = demonstrable in their own service area. 4-finding brief. Teach-first ask citing SCE's own LiDAR/ML posture.
concept · not affiliated Open the review →Chevron
Offset credibility is Chevron's weakest public flank, it retired ~5.8M carbon credits (2020-2022) and independent research publicly disputes the integrity of the bulk (attributed as that party's claim), yet a retired credit is a registry serial with no traceable, verifiable proof-of-work behind it. It is also the least agent-legible of the four: the Akamai edge 403s every non-browser user-agent (AI/ESG bots blocked at the door), robots.txt names only social crawlers, and behind a browser the agent well-knowns are all true 404 with 0 ld+json on the Sitecore homepage. EWP proposed as the proof-of-work-behind-the-credit layer, legacy well-site remediation as hazard-gated, photo+GPS+signature-proofed packets, with a REAL on-chain Klima retirement bound to each packet (the one engagement in this batch that leans on the live retire rail). 4-finding brief. Teach-first ask citing Chevron's own New-Energies/AI investment.
concept · not affiliated Open the review →WM (Waste Management)
North America's largest environmental-services network, ~257 active landfills, ~342 transfer stations, the largest recycler and collection fleet, the leader in landfill-gas-to-energy, ~$25B revenue, runs hundreds of thousands of recurring proof-of-completion events a year (methane/air surveys, groundwater/PFAS screens, stormwater outfalls, closed-landfill restoration, field-crew safety) with no per-site machine-verifiable proof layer. Its security baseline is reasonable (HSTS preload, CSP frame-ancestors, a fully-open robots.txt, a well-formed 9,076-URL sitemap), but the homepage is a Next.js loading shell serving 0 ld+json / 0 og tags, llms.txt is a true 404, the whole /.well-known/ tree returns a 403 wall (agents fail closed, not merely 404, contradicting the open robots.txt), and no security.txt is readable anywhere. EWP proposed as the per-site proof layer, methane/monitoring + closed-landfill restoration packets, safety-gated and settled on-chain, the use-of-proceeds argument settled at the site level. The strongest lane-for-lane EWP fit in the portfolio. 4-finding brief. Teach-first ask citing WM's own methane-analytics (Blue Sky/AirLogic) bet.
concept · not affiliated Open the review →Regen Network
Not a cold target, operator is a Regen Tokenomics WG steward with two open PRs on Regen's own repos (agentic-tokenomics#108, regen-compute#139, re-verified live). Regen Compute ships a genuinely agent-ready surface: 3 real discovery manifests (MCP server-card, A2A card, multi-step workflow manifest), a full OpenAPI spec, and a standards-conformant x402 flow, ahead of nearly every target in this portfolio. But Regen's own OpenAPI spec documents that POST /retire can silently fall back to a human-completed marketplace link when no wallet/liquidity; app.regen.network (the actual marketplace) has zero agent-discovery surface (soft-200 SPA shell on every path incl. a real project + a control); www.regen.network and the Webflow-hosted registry both ship a real-but-empty robots.txt. Underneath it all: no bounded, funded, proof-verified, on-chain-settled layer for the field labor (MRV monitoring, planting checks, boundary surveys) a credit requires before issuance, EWP proposed as that missing labor leg. 6-finding evidence-led brief + relationship context.
EWP-integration concept Open the review →Klima Protocol
EcoWealth is a LIVE INTEGRATOR here, not a cold read: our retire rail calls x402.klimalabs.com/api and the Retirement Aggregator on Base, gasless. It ran hourly until retirements are stopped by operator order as of 2026-08-04. Klima/Carbonmark ship one of the best-documented agent stacks in the portfolio (real llms.txt x2, full GitBook llms-full.txt exports, a real 84 KB OpenAPI spec, a published Claude-Code skill plugin on GitHub), credited in full. But the standard discovery paths lie: klimaprotocol.com soft-200s every well-known path (agent-card.json, security.txt, x402) with byte-identical homepage bytes; carbonmark.com soft-400s the same; app.klimaprotocol.com (where retirement actually happens) true-404s everything; the branded 'x402 endpoint' has no manifest at .well-known/x402; and carbon.fyi (the new discourse-fixing news product) ships no robots.txt, no llms.txt, no feed at all. 6 findings, curl-reproducible, credit-first.
EWP-integration concept Open the review →bridge.eco (ecoBridge), cross-chain payment/retirement bridge for Regen Network ecological credits, and the ecoToken landscape (ecotoken.earth, app/eclipse/aptos.ecotoken.earth, scan.ecotoken.earth, RegenAtlas.xyz, CCEP/ecoLedger protocol)
Warm relationship, not a cold read, EcoWealth's regen-compute MCP integration already ships browse_ecobridge_tokens and retire_via_ecobridge; ecoBridge's own CEO (github ecoTokenJames) has a direct commit on regen-network/regen-compute, the same repo EcoWealth's operator (brawlaphant) has 5 commits on; operator holds a warm contact for bridge.eco. ecoBridge is genuinely capable: a Squid Router-powered any-asset swap across 70+ chains, a real open/CORS/no-auth registry API (api.bridge.eco/registry) with an LLM-ready quality+sales-enablement object per project, and an infrastructure-deep Regen tie ($REGEN as CCEP/ecoLedger's own governance token since Oct 2024, ecoLedger data posted to Regen Ledger, a live ecoBridge-run validator on Regen Network). But its own 'AI' tab (ecoAgent) is a proprietary human chatbot inside a closed SPA with zero machine-discoverable manifest anywhere (bridge.eco soft-200s every well-known agent path with an identical 465B shell; ecotoken.earth true-404s the same paths; 0 mentions of 'agent'/'MCP' across 28KB of ecoBridge's own docs), the only real agent front door to ecoBridge is Regen's own MCP server. The 'all Regen credit types' claim doesn't match the live registry: 8 Carbon + 1 correctly-labeled Terrasos(BT01) + 1 project mislabeled 'Carbon' that is actually KSH01 (batch-code verified against Regen's own project URI) + zero Marine Biodiversity/Umbrella Species. Traffic-readiness audit (operator hypothesis) evidenced: bridge.eco transfers 28.6MB with a real broken-import console error (Stellar wallet init); ecotoken.earth throws a real page error (Bootstrap requires jQuery); neither domain ships a single Open Graph/Twitter Card tag; app.ecotoken.earth loads in 11.5s at 35.7MB. Credited where solid: scan.ecotoken.earth and the core bridge.eco flows render clean; ecotoken.earth's homepage passes the 5-second test clearly even though bridge.eco's own app does not yet.
EWP-integration concept Open the review →Oimpact
Warm-relationship concept, existing operator contact (not exercised). Oimpact (canonical domain oimpact.ai; 0impact.ai 301-redirects to it; app.oimpact.ai's own product titles itself "OpenImpact"; operating team is ecoToken, GitHub org ecoWarriors) launched literally hours before this review, its own /leaderboard says "JUST LAUNCHED... stats are zero today." It ships one genuinely real agent-facing door: api.oimpact.ai self-describes in one GET and its two documented GET endpoints return honest 401s, not soft-200s. But that same self-description points integrators to a GitHub docs URL that's a true 404 (confirmed via the public GitHub API), the marketing domain soft-200s every standard agent-discovery path (robots.txt/llms.txt/agent-card.json/sitemap.xml all return the identical SPA shell, differing from a nonsense control by exactly one reflected og:url string), and oimpact.ai has no SPF/DMARC at all despite live Google-Workspace mail. Credit where due: a genuinely excellent, self-serve /brandkit (exact hex palette, type system, voice guide) supplied every color used in this brief's demo. Underneath the RWI mint sits a real, live proof-of-outcome layer one hop away, 0G Labs' own climate.0g.ai, citing 4,000+ tCO2e retired via Regen Network's City Forest Credits, but nothing in Oimpact's own stack funds or verifies the field labor a credit needs before it exists; EWP proposed as that missing evidence layer. 6 findings, curl-reproducible.
concept · not affiliated Open the review →Sylvera
Sylvera co-chairs CDOP, an open carbon-data standard whose Unit_Description schema already has a field (retirement_beneficiary, id 141) matching EcoWealth's own beneficiary-never-sender law almost verbatim. We built a sample EWP settlement receipt in that shape and validated it against CDOP's own real, published schema tooling (check-jsonschema) -- catching and fixing one real formatting error along the way -- proving the fit rather than just claiming it. Sylvera's own site carries no agent-discovery surface, but honestly (true 404s, not soft-catches); its Assessment API docs render nothing without JavaScript; pricing is contact-sales-only above a free tier. 5 findings, credit-first, one of them a real validator run.
concept · not affiliated Open the review →Carbon Gap
NOT AFFILIATED, no contact made, no relationship claimed. Carbon Gap's own 20 May 2026 demand-survey brief found that among 25 senior sustainability leaders, the plurality reason CDR purchases stall isn't cost or policy per se, it's quality, defined verbatim as 'real, durable, measurable, independently verified removals' (11/25, 44%, ahead of cost 7/25 and policy certainty 5/25). That is close to a literal description of an EWP receipt: a funded, photo+GPS+signature-proof, on-chain-settled work packet, independently re-checkable by anyone. Genuine credit due: the CDR Policy Tracker (~40 policies, 30+ countries) and Funding Database (700+ opportunities) are real, open, no-paywall tools, and carbongap.org runs one of the more honest agent-discovery surfaces reviewed in this portfolio (real llms.txt + robots.txt, true 404s elsewhere, no soft-200 desert). Courtesy finding: two homepage-linked subdomains (guide., care.) fail DNS entirely. Opportunity: the CRCF carbon-farming certification methodology adopted 10 Jul 2026 leaves peatland-rewetting MRV as a zero-match, one-week-old, zero-competition gap in our own library, while soil-carbon and afforestation defs already align closely. One figure from the original recon pass ('65% cite clearer standards') did not verify against the source survey and was corrected to the confirmed 44% Quality figure, disclosed on every page.
concept · not affiliated Open the review →Regreener
Not affiliated, no relationship, no contact made. Regreener (regreener.earth, ~10 people, B Corp since 2024, bootstrapped + a EUR 350k 2022 angel round) sells trust in a curated shortlist, not credits at scale: a published 5-category, 100+/200+-data-point quality framework vets carbon projects, survivors get published as SEO-driven '5 Best X of 2026' buyer's guides, sold onward to ~240 clients with no public pricing and no API. The site is honestly built where it matters, real robots.txt/sitemap, correctly configured SPF/DMARC, but every agent-discovery convention (llms.txt, agent-card.json, .well-known/x402) is a true, control-matched 404: a genuine absence, not a lie. The vetting itself is real but private and unrepeatable by a buyer; the domain's only blockchain reference is one hedged FAQ sentence ('...where possible') on a single product page. Regreener's own sourced pricing data (EUR 12.3B 2025 offtake at ~EUR 180/credit vs. EUR 1.04B spot at ~EUR 6.10/credit, a ~30x premium) argues EWP work should be sold as standing packet commitments to this buyer class, not one-off x402 calls. Proposed complement, not replacement: EWP's on-chain settlement receipt as the re-checkable evidence layer underneath a Regreener-curated pick. 6 findings + 2 context notes, all curl/dig/fetch-reproducible.
concept · not affiliated Open the review →CDR.fyi
NOT AFFILIATED, and structurally out of scope for their tracker, stated plainly, not pitched around. CDR.fyi's own public methodology page (docs.cdr.fyi/methodology, live) says verbatim it aggregates carbon 'stored for +100 years'; EcoWealth's retirements settle a Regen-CFC nature-based (sub-100-year) class at 5-cent micro-retirement scale, clearing neither the durability nor the whole-tonne-schema gate. That said, CDR.fyi runs one of the cleanest agent-facing surfaces reviewed in this portfolio: a genuinely excellent llms.txt (200, 3827B) with explicit AI-citation governance ('AI systems should note that premium Portal content is behind authentication and should not be cited as freely accessible'); true 404s (byte-identical to a nonsense control, 28695B) at every non-existent well-known path, the opposite of the soft-200/soft-400 catch-alls found on two other protocol/marketplace sites reviewed in this portfolio; and a genuinely honest, versioned API gate (api.cdr.fyi/v1/orders returns a real 401 'Invalid write key' with zero credentials, not a fake success). The one narrow gap: the data llms.txt itself calls free (leaderboard/market-update figures) still has no machine-readable schema, an agent has to scrape rendered JSON to reach it. The package proposes two small, honest additions (a public-only stats manifest; an attributed CDR/OPIS price-context object hosted on EcoWealth's own domain, not CDR.fyi's) and a give-to-get vs. free-proof transparency-model comparison, no pitch for tracker inclusion anywhere.
concept · not affiliated Open the review →BeZero Carbon
NOT AFFILIATED, an independent concept note, not a cold audit and not a relationship. BeZero rates registry-issued carbon credits AAA-D against six named risk factors (Additionality, Over-crediting, Leakage, Non-permanence, Information, Policy, corrected against an earlier internal draft that had mis-cited 'perverse incentives'), backed by 80+ scientists and engagement with 500+ project developers, with research showing each ratings notch adds ~30% average price premium. This note applies the same six factors, credited throughout, to a single real, already-defined EWP work packet (a Los Angeles street-tree planting) one layer beneath where a credit is ever rated, honestly marking which factors proof-based labor evidence helps with (additionality, information) and which it plainly doesn't (non-permanence, policy, tCO2e quantification). Also corrects two smaller drifts from the source recon: the public listings '+680 results' table is BeZero's own explicitly-labeled illustrative/fictitious demo data, not a live count; and the developer-engagement figure is 'over 500' projects, not 400+.
concept · not affiliated Open the review →Carbonfuture
NOT AFFILIATED, independent concept. Carbonfuture runs genuinely institutional-grade dMRV (9-criteria due diligence, 100+ data points, accredited-lab batch sampling, VVB sign-off) and real 2025 market scale (1.5Mt+ facilitated, the Microsoft x Exomad Green 1.24Mt megatonne biochar deal, a SIX Group-led Series A2), credited in full; this is a different, harder proof problem than EWP solves. The honest gap: EWP's photo+GPS+signature proof settles a single bounded task in one transaction; Carbonfuture's dMRV proves carbon quantity and durability for populations of tonnes. The pitch is deliberately narrow, EWP as a last-mile field-capture layer feeding batch-level dossiers upward, plus packet-sized funded labor (feedstock collection, invasive clearing, site prep) around the 'missing middle' plants Carbonfuture's own thesis names as under-financed, never a competing standard. Self-critique included: EWP's own proof-hash convention isn't yet wired to a public resolver, the same bar Isometric's registry already clears. Fresh agent-discovery check (recon didn't cover it): carbonfuture.earth ships real robots.txt/sitemap.xml but true-404s every llms.txt/agent-card/x402 path (byte-identical to a nonsense control, honest, not a soft-200 fake). 6 evidenced findings, credit-first.
concept · not affiliated Open the review →Frontier
Unaffiliated concept package, EcoWealth is not a Frontier member, buyer, supplier, applicant, or Symbiosis participant. Frontier is the most price-transparent buyer in durable carbon removal reviewed in this portfolio: $694M contracted / $1.8B committed, 18 named offtakes priced $214-$989/t (verified live), and a CC0-licensed public-domain offtake-agreement template on GitHub (license + activity independently confirmed). This package corrects a real public-record conflation, Microsoft and Amazon are NOT Frontier members, confirmed directly against Frontier's own member-logo markup (upgrading a prior recon's search-only caveat to a primary-source check), credits Anthropic's and Salesforce's 2026 arrival, flags that frontierclimate.com itself ships zero agent-discovery surface (all true 404s, control-matched), and demos Frontier's own published offtake mechanics (pay-on-delivery, conditions precedent, grace periods, buyer-side MFN, ROFO) running at EWP packet scale as a labeled concept. 6 findings, curl-reproducible.
concept · not affiliated Open the review →Charm Industrial
Unaffiliated lean recon. Holds Frontier's first-ever offtake ($53.0M / 112,000t / $473/t, 2024–2030), verified live against frontierclimate.com/writing/first-offtake. Genuinely credit-worthy: charmindustrial.com/llms.txt returns a real 200 with 6,302 bytes of hand-authored content, rare in this portfolio. Gap: no /.well-known/agent-card.json (true 404, control-matched). EWP field-labor angle (concept only): feedstock collection/hauling and injection-site monitoring/restoration.
concept · not affiliated Open the review →Lithos Carbon
Unaffiliated lean recon. Holds Frontier's first-ever enhanced-weathering offtake ($57.1M / 154,240t / $370/t, 2024–2028), verified live against frontierclimate.com/writing/lithos. Genuinely distinctive operating model: basalt spreading happens on active partner farmland, not a fixed plant. Gap: llms.txt and /.well-known/agent-card.json both true-404, control-matched, zero agent-discovery surface of any kind. EWP field-labor angle (concept only): farm-by-farm basalt spreading and soil-sampling verification.
concept · not affiliated Open the review →Vaulted Deep
Unaffiliated lean recon. Holds a real Frontier offtake ($58.3M / 152,480t / $382/t, 2024–2027), verified live against frontierclimate.com/writing/vaulted. Credit: every unmapped path (agent-card.json, a nonsense control) returns a byte-identical 209,646-byte true 404, no soft-200 deception found. Gap: no llms.txt or any agent-discovery manifest beyond a real, minimal robots.txt (116B). EWP field-labor angle (concept only): waste-feedstock hauling, well-pad monitoring, and land restoration.
concept · not affiliated Open the review →Stockholm Exergi
Unaffiliated lean recon. Holds a real Frontier BECCS offtake ($48.6M / ~179,998t provisional / ~$270/t, 2028–2030), verified live against frontierclimate.com/writing/exergi. Honest caveat carried forward: the original 2024 announcement left tonnage TBD pending a Swedish reverse auction; the live /portfolio figure should be treated as provisional until a primary source confirms the auction closed at that volume. Gap: llms.txt and agent-card.json both true-404, control-matched, zero agent-discovery surface. EWP field-labor angle (concept only): biomass feedstock handling and capture-equipment monitoring support.
concept · not affiliated Open the review →280 Earth
Unaffiliated lean recon. Holds a real Frontier DAC offtake ($40.0M / 61,571t / $650/t, ~2025–2030), verified live against frontierclimate.com/writing/280-earth. Real gap, confirmed twice with a standard browser UA and request spacing to rule out Cloudflare bot-challenge noise: /.well-known/agent-card.json and a nonsense-path control both return HTTP 200 with the full 140,645-byte homepage, while /llms.txt correctly true-404s at a small, distinct WordPress 404 page, the soft-200 is specific to the extensionless/well-known path space, not universal, but it's real and reproducible. EWP field-labor angle (concept only): DAC-array site prep, monitoring, and land restoration.
concept · not affiliated Open the review →Heirloom
Unaffiliated lean recon. Holds the highest $/tonne deal in Frontier's entire named-offtake portfolio ($26.6M / 26,900t / $989/t, 2023 vintage), verified live against frontierclimate.com/writing/carboncapture-heirloom. Credit: llms.txt (21,481B) and /.well-known/agent-card.json (88B) both true-404, each exactly matching its own path-matched control, two different but both-honest 404 responders, no soft-200 deception anywhere. Gap: no agent-discovery manifest beyond a minimal, real robots.txt (51B). EWP field-labor angle (concept only): limestone material handling and site restoration.
concept · not affiliated Open the review →CarbonCapture Inc.
Unaffiliated lean recon. Holds a real Frontier offtake ($20.0M / 45,500t / $440/t, 2025–2028), announced the same week as Heirloom's deal (combined $47M writeup), verified live against frontierclimate.com/writing/carboncapture-heirloom. Credit: llms.txt and /.well-known/agent-card.json both true-404 at a byte-identical 24,449 bytes, exactly matching a nonsense control, fully consistent, no soft-200 anywhere. Gap: no agent-discovery manifest beyond a real, present robots.txt (235B). EWP field-labor angle (concept only): modular-array site-pad prep and monitoring rounds.
concept · not affiliated Open the review →Puro.earth
NOT AFFILIATED, independent concept. Puro.earth is a standard-setter AND registry operator for durable carbon removal, a live, daily-updated registry (1.78M+ CORCs issued, 1.04M retired, 117 certified projects) across 8 certified methodologies (biochar flagship, 200-1,000+ yr durability classes). A materially harder proof problem than EWP's task-completion proof: Puro requires calibrated weighbridges, accredited-lab Corg/H-ratio analysis, and VVB sign-off before a CORC issues. The honest pitch is deliberately narrow, EWP as a last-mile field-capture layer feeding a batch dossier upward, never a parallel CORC issuer. Fresh agent-discovery check: real robots.txt but llms.txt/agent-card/openapi all true-404 (byte-identical to a nonsense control), a gap, not a fake. 4 evidenced findings, credit-first.
concept · not affiliated Open the review →Isometric
NOT AFFILIATED, independent concept. Isometric is an AI-assisted certification platform + public registry (145,260 certificates issued, 39,535 retired) that already publishes "full provenance and data behind each certificate", the exact transparency bar this six-entity batch measures against, and one EWP's own proof-hash convention (built, not yet wired to a public resolver) hasn't fully cleared. $40M raised; Google/Boeing/Veolia/Anglo American/Microsoft as customers; Biochar Production and Storage protocol v1.0 ICVCM-approved August 2025. Evidence vocabulary (sensor/video/satellite/LiDAR/lab results) is genuinely broader than EWP's photo+GPS+signature. Fresh agent-discovery check: real robots.txt/sitemap but true-404 llms.txt/agent-card family, the same Webflow-platform pattern independently confirmed on a second target in this portfolio (Carbonfuture was the first). 4 evidenced findings, credit-first.
concept · not affiliated Open the review →Exomad Green
NOT AFFILIATED, independent concept. Exomad Green is a Bolivia-based biochar producer claiming the largest CDR delivery volume industry-wide (1.4Mt removed to date, ~24.5% of all durable removals), operating two Sylvera/BeZero-rated plants (Concepción 140kt/yr, Riberalta 120kt/yr) with a third under construction, underneath a 1.24Mt/10yr Microsoft megadeal. The standout finding, credited first: Exomad is the ONLY target in this six-entity batch, and rare across the whole EcoWealth portfolio, with a real, working AI-agent integration (a live Wix MCP endpoint at /_api/mcp, documented in a genuine llms.txt), not just a static discovery file. EWP's honest role is provable field labor around the production sites (feedstock collection, farmer distribution days, site prep), never a carbon-credit claim. 4 evidenced findings, credit-first.
concept · not affiliated Open the review →Biomass Projects
NOT AFFILIATED, independent concept. Biomass Projects is developing the Mardie char project in the Pilbara, WA, converting invasive mesquite (300,000+ ha regional infestation draining ~11B m³ groundwater/yr) into biochar, targeting 500,000 t CO2/yr by 2028 (full build 10.5Mt biomass / 10.2Mt CO2 over 20 years), partnered directly with the Wirrawandi Aboriginal Corporation (Mardudhunera and Yaburara Traditional Owners). Sits precisely in Carbonfuture's own named "missing middle" financing gap, a plant-level, multi-million-dollar CAPEX gap EWP's sub-dollar packets cannot fill directly. EWP's honest role is packet-sized field labor around the site (mesquite clearing, site prep, groundwater monitoring), coordinated with, never substituting for, the Traditional Owners' own workforce. Fresh agent-discovery check: real robots.txt but llms.txt/agent-card/openapi all true-404 (byte-identical to a nonsense control), consistent with an early-stage developer site. 4 evidenced findings, credit-first.
concept · not affiliated Open the review →NextGen
NOT AFFILIATED, independent concept. NextGen is a South Pole x Mitsubishi Corporation JV buyers' club aggregating real demand (Swiss Re, BCG, LGT, Mitsui OSK Lines, UBS, later Mizuho) at a stated $200/tonne target price, with an explicit anti-double-counting commitment ("all CDRs... retired... in a public registry") backed by named independent dMRV (Carbonfuture) and a recent direct agreement with Exomad Green. The five-pathway portfolio (BiCRS/DACS/Enhanced Weathering/High-temp Biochar/Product Mineralization) carries zero disclosed percentage breakdown, and the retirement promise reads as prose, not a queryable feed. EWP's settlement-is-registry-entry pattern offered only as an illustration of machine-verifiable buyer transparency, not a proposal. Fresh agent-discovery check found a stricter failure mode than the rest of this batch: true-404 llms.txt family, but a flat 403 Forbidden on the entire /.well-known/* namespace. 4 evidenced findings, credit-first.
concept · not affiliated Open the review →OPIS
NOT AFFILIATED, independent concept. OPIS (a Dow Jones company) is a regulated, IOSCO-audited price reporting agency whose RIN/LCFS/CCA carbon-market benchmarks genuinely settle real CME Group futures contracts, credited in full as load-bearing market infrastructure, not a target EWP competes with or could replace. Two gaps named plainly: (1) OPIS's own robots.txt specifically Disallows crawling the exact three IOSCO PRA Principles compliance-audit PDFs (2022-2024) that are its own trust-verification evidence, likely an unintentional generic-rule side-effect, not deliberate, but ironic given the "verification" framing; (2) no agent-discovery manifest (llms.txt/agent-card/openapi all true-404) and no free/agent-queryable pricing feed, the real numbers sit behind subscription products (RackPro, Context), a standard Bloomberg/Platts-style paywall model, not an open public good. EWP's own on-chain settlement price is offered only as a small illustration of open, machine-readable pricing infrastructure at a far smaller scale, not a claim of competing with OPIS. 4 evidenced findings, credit-first.
concept · not affiliated Open the review →Verra
Verra operates the world's largest voluntary carbon program (VCS), 2,579+ registered projects, 132+ countries, 1.3B+ tonnes reduced/removed cumulative, all re-verified directly from Verra's own program page. Verra was also the last of the major programs to clear ICVCM's CCP-eligibility, and only after required procedural changes, a real credit, not a rubber stamp. Agent discovery is an honest, two-layer gap: real robots.txt/sitemap infrastructure, but llms.txt/security.txt/ai.txt/agent-card.json all true-404 (confirmed against a same-host nonsense-path control, not a soft-catch). The core proposal: Verra's VVBs verify aggregated monitoring reports; EWP's photo+GPS+signature+on-chain-settlement receipt is offered as a candidate evidence layer one level beneath that report, never a claim to issue VCUs, quantify tCO2e, or compete with Verra's registry. 3 evidenced findings plus an honest EWP-intersection section, all curl-reproducible.
concept · not affiliated Open the review →Gold Standard
Gold Standard for the Global Goals is the only major registry built explicitly around multi-outcome impact (GHG + health aDALYs + water + waste + more) rather than carbon alone, per its own SDG Impact Quantification page, 'multiple methodologies may be applied in one project.' ICVCM has specifically confirmed Gold Standard's Clean Cooking methodologies as CCP-Eligible, a named methodology-level approval. Agent discovery is a real, honest gap: robots.txt and sitemap are genuine, but llms.txt/security.txt/ai.txt/agent-card.json all true-404 with an identical branded template, confirmed against a same-host control. The core proposal: Gold Standard's multi-outcome quantification happens at the aggregated impact-report level; EWP's per-task photo+GPS+signature+on-chain-settlement receipt is offered as a candidate evidence layer beneath it, mirroring the same multi-outcome instinct (EWP's own three-ladder library) at packet scale. 3 evidenced findings plus an honest EWP-intersection section.
concept · not affiliated Open the review →Plan Vivo
Plan Vivo is the oldest smallholder/community-focused carbon standard (founded 1994), distinguished by requiring a defined majority of project revenue (historically >=60%) flow directly to communities rather than developers. Its new PV Climate V5 standard, per Calyx Global's independent case study, splits credits into fPVC (future/ex-ante)/rPVC (reported, unverified)/vPVC (verified) and explicitly bars retirement of anything but a vPVC, a real, recent integrity tightening. Agent discovery shows a genuine mixed picture: Plan Vivo actually ships a real security.txt (rare in this batch) but its Contact field is an unfilled placeholder ([email protected]); llms.txt/agent-card.json/ai.txt all true-404 identically. The core proposal: vPVC verification still happens at monitoring-report/registry scale; EWP's per-task photo+GPS+signature+on-chain-settlement receipt is offered as a candidate primary-evidence layer beneath that report, alongside a shared instinct (pay/credit the people who did the work) with Plan Vivo's own community-revenue model. 3 evidenced findings plus an honest EWP-intersection section.
concept · not affiliated Open the review →ICVCM
ICVCM owns the Core Carbon Principles (CCP) label, 10 principles across Governance/Emissions Impact/Sustainable Impact buckets, the market's emerging cross-registry quality bar. Its Governing Board has real teeth: independent coverage states 40 of the first 65 assessed methodologies were approved and 25 were rejected outright. Nine programs are now CCP-Eligible (ACR, ART TREES, CAR, Equitable Earth, Gold Standard, GSS, Isometric, Puro.Earth, Rainbow, and, most recently, after required changes, Verra). Agent discovery is a real, honest gap: robots.txt is genuine, but llms.txt/security.txt/ai.txt/agent-card.json all true-404 identically, confirmed against a same-host control. The core finding is about scope, not rigor: ICVCM assesses methodologies and programs, never individual projects, CarbonPlan's independent finding (6 of 14 CCP-labelled landfill-gas projects failing an additionality test) proves that gap is real, and it's exactly the layer EWP's per-task proof receipt is built to sit beneath, without asking ICVCM to change its own mandate at all. 3 evidenced findings plus an honest EWP-intersection section.
concept · not affiliated Open the review →SBTi
SBTi's Corporate Net-Zero Standard V2.0 is final (published 2026-06-11) and effective 2027-01-31, with a dated, real compulsory-demand clock: from 2035, large companies must cover a rising share of Scope 1-3 emissions (1% to 100% by net-zero year) via eligible carbon removals, at least 10% of which (for long-lived GHGs) must be long-lived (century-plus) removals. SBTi's own definitional bar for a qualifying removal, ex-post, independently third-party assured, permanently retired, no double counting, no forward-looking promise, is precisely the shape of an EWP proof-approve-settle-retire receipt. Agent discovery is a real, honest gap: robots.txt/sitemap are genuine but minimal, while llms.txt/security.txt/ai.txt/agent-card.json all true-404 to an oddly-titled 'search' page, confirmed against a same-host control. This is not a claim that EWP packets qualify as SBTi-eligible removals today, durability and tCO2e quantification remain separate, unbuilt questions, it's a note that the evidentiary discipline the 2035 mandate will require already has a working implementation, years ahead of the deadline. 3 evidenced findings plus an honest EWP-intersection section.
concept · not affiliated Open the review →Stellar Green
Stellar Green Corp. (ステラーグリーン), a 100%-owned SB Players subsidiary founded 2024-05-09, supports forest-carbon-credit creation under Japan's domestic J-Credit Scheme across roughly 5,000 hectares of municipal-managed and private forest, built through direct one-relationship-at-a-time partnerships with multiple Hokkaido municipalities, a forestry cooperative, and a regional bank. It has partnered with Sylvera specifically to make that credit quality visible to corporate buyers, a named, credit-first quality initiative, not just a volume play. Agent discovery is the bluntest gap in this batch: no robots.txt at all, every agent path true-404ing byte-identically to a control. The core proposal: every hectare Stellar Green brings into the J-Credit Scheme still needs field verification before a credit issues; EWP's photo+GPS+signature+on-chain-settlement receipt is offered as a candidate primary-evidence layer for that field work, expressible in the CDOP vocabulary Sylvera already uses internationally (per this portfolio's own Sylvera brief), with zero claim of any present relationship, registration, or J-Credit Scheme presence for EWP today. 3 evidenced findings plus an honest EWP-intersection section.
concept · not affiliated Open the review →Offsetra
Offsetra built carbon.fyi in 2019-2021, one of the first tools estimating an Ethereum wallet's carbon footprint from mining-pool geolocation and transaction timing, open-sourced (GitHub repos still live, confirmed 200 as of this recon) and later handed to new stewards: carbon.fyi's own /about page states plainly 'This website is maintained by Klima Protocol and Carbonmark' while still crediting Offsetra by name and linking to the original tool. Offsetra itself continues independently, reporting 1,054,243 tonnes CO2e offset to date via corporate/bulk offset retail and blockchain/NFT emissions consulting. Agent discovery is the starkest 'nothing at all' case in this batch: both offsetra.com and the current carbon.fyi true-404 every path including robots.txt itself, byte-identical to a nonsense-path control on each domain. This brief is framed differently from the others in this batch, it's prior-art credit and lineage, not a gap audit: EWP's settlement-and-retirement receipt is offered as the same underlying instinct (making a specific on-chain action's real-world impact legible and actionable) continued forward instead of backward, with zero claim of relationship, adoption, or comparison of scale.
concept · not affiliated Open the review →Autodesk
Autodesk's own operational sustainability numbers (100% renewable energy, SBTi-validated targets) and its 2023 Frontier Climate membership are real and credited in full; the honest gap is that neither Autodesk's design software nor Frontier's removal ledger proves the physical field work that makes a modeled building real, EWP's proof packet is proposed as a complementary receipt layer, never a competing claim.
concept · not affiliated Open the review →H&M Group
H&M Group's SBTi-verified 2030/2040 climate targets and 2023 Frontier Climate membership are real and credited in full, with the 2040 target explicitly naming permanent carbon removals as its backstop; the honest gap is that the repair/resale/take-back labor that reduces how much needs offsetting has no receipt layer of its own, EWP's proof packet is proposed as that layer, never a claim of contact or partnership.
concept · not affiliated Open the review →SIX Group
SIX Group is credited for real, working agent-discovery infrastructure (llms.txt, security.txt) and for a genuine dual buyer+investor role in carbon-removal trust infrastructure via Carbonfuture; one precise correction distinguishes SIX's actual role from an over-broad Article 6 claim. EWP's own settlement-as-registry-entry design is proposed as a structural parallel worth SIX's attention, never a claim of contact or partnership.
concept · not affiliated Open the review →City of Stockholm
The City of Stockholm's 750,000-tonne BECCS removal purchase, its 50% Stockholm Exergi ownership, and its numbered Climate Action Plan 2030 are all real and credited in full; the honest gap is that no resident-checkable receipt exists for the small, local, everyday ecological labor (parks, waterways, urban forestry) that sits between the industrial removal number and the street, EWP's SDG-anchored 'Restore' ladder is proposed as that layer, never a claim of contact or procurement relationship.
concept · not affiliated Open the review →Microsoft
Unaffiliated concept package, EcoWealth is not a Microsoft vendor, partner, supplier, or coalition co-member. Microsoft's own newsroom confirms 45 million tonnes of carbon dioxide removal contracted in fiscal year 2025 (2x FY2024, 9x FY2023) across 21 suppliers and 10 countries, the largest single-company removal footprint reviewed in this portfolio, run entirely outside the Frontier coalition it is often wrongly conflated with (independently confirmed absent from Frontier's own member-logo markup in a prior recon pass). Microsoft instead co-founded the separate Symbiosis Coalition (with Google, Meta, Salesforce) for nature-based removal. This package credits that scale, flags that microsoft.com ships zero agent-discovery surface (all true 404s, control-matched), and notes the honest EWP intersection: Microsoft's own Copilot/Azure AI Foundry agent platforms could consume x402-priced, receipt-backed EWP packets directly as the visible local complement to its far-away tonnage. 4 findings, curl/press-reproducible.
concept · not affiliated Open the review →Stripe
Unaffiliated concept package, EcoWealth is not a Stripe partner, Stripe Climate customer, or Frontier co-member. Stripe founded and owns Frontier's carbon-removal AMC LLCs and runs its own Stripe Climate merchant-funded removal program (climate.stripe.com: '$1B+ of permanent carbon removal by 2030'). Separately, and more relevantly to EWP, Stripe ships a real, substantial llms.txt (65KB/379 lines) and co-launched the x402 open HTTP-402 USDC-settlement standard with Coinbase (Feb 2026), the identical protocol family EWP's own paid pages already settle through. The one gap: no .well-known/agent-card.json despite naming an 'Agentic Commerce Protocol.' 4 findings, curl/press-reproducible, credit-first, no claimed relationship.
concept · not affiliated Open the review →Shopify
Unaffiliated concept package, EcoWealth is not a Shopify merchant, app-store partner, or Frontier co-member. Shopify co-founded Frontier (2022) and separately runs its own Sustainability Fund ($94M cumulative contracted across 54 suppliers, 193,000 tonnes contracted / 16,000 delivered per its 2024 Climate Report) plus the merchant-facing Planet checkout app (22M+ orders funded cumulatively, 56,000+ tonnes during BFCM 2023 alone), real, credited in full. This package's one finding: shopify.com's own real, live llms.txt (200, 861 B) never mentions any of it, and .well-known/agent-card.json 404s at a suspiciously small 9 bytes versus the ~57KB branded 404 every other unknown path gets, suggesting a different, unwired edge layer. 4 findings, curl-reproducible, credit-first, no claimed relationship.
concept · not affiliated Open the review →McKinsey Sustainability
Unaffiliated concept package, EcoWealth is not a McKinsey client, Frontier co-member, or engagement partner. McKinsey Sustainability's own blog confirms it co-founded Frontier in 2022 (with Stripe, Alphabet, Shopify) and has committed to shifting its own offsetting entirely to carbon removals by 2030 alongside SBTi-validated 2025/2050 targets, both credited in full. The distinctive finding: mckinsey.com's edge rejected every plain, passive request attempted this session, including robots.txt (TLS handshake stall) and a sustainability sub-page (403), the single most locked-down domain probed across this six-entity batch; McKinsey's own facts here are sourced via search-index corroboration of its own pages, not a live fetch. Its nonprofit-facing mckinsey.org loaded fine and shares the same real design system (McKinsey Sans + Bower, confirmed live), showing the block is domain-specific. 4 findings, credit-first, no claimed relationship.
concept · not affiliated Open the review →Salesforce
Unaffiliated concept package, EcoWealth is not a Salesforce customer, ISV partner, or coalition co-member. Salesforce joined Frontier's June 2026 Growth AMC round (board seat via Jamila Yamani, confirmed in a prior EcoWealth recon), co-founded the Symbiosis Coalition, and by public reporting is the only company in all three of Frontier/Symbiosis/LEAF, credited in full. salesforce.com also ships the largest real llms.txt found across this six-entity batch (1MB+, post-redirect). One structural observation: the domain's Akamai edge redirects everything, including a nonsense control, and sets a cookie literally named sfdc_agent_cohort=qualified, while .well-known/agent-card.json still true-404s. Agentforce for Net Zero Cloud (Salesforce's own AI agent for sustainability reporting) is a concrete, shipped example of the kind of product EWP's receipt-backed proof data could plausibly feed. 4 findings, curl/press-reproducible, credit-first, no claimed relationship.
concept · not affiliated Open the review →Workday
Unaffiliated concept package, EcoWealth is not a Workday customer, partner, Frontier co-member, or Workday Build developer. Workday's own blog confirms it joined Frontier in 2023 as part of a combined $100M AMC commitment with JPMorgan and H&M, and separately runs a real internal program (100% renewable-matched electricity, a carbon-neutral cloud for its whole customer base, net-zero offices/data centers/travel, a $1M mangrove-reforestation investment, and a stated 2031 target to convert its whole credit portfolio to removals), credited in full. workday.com ships zero agent-discovery surface (all true 404s, control-matched within 6 bytes). The sharper product-level note: Workday's own Illuminate AI agents are metered by 'Workday Flex Credits,' a transparent, consumption-based model structurally identical to the x402 per-packet settlement EWP already runs for ecological labor. 4 findings, curl/press-reproducible, credit-first, no claimed relationship.
concept · not affiliated Open the review →Dr. Alexander D. Wissner-Gross
An independent, not-affiliated engagement review of alexwg.org and Dr. Wissner-Gross's corpus (24 publications, 26 patents, 128 distinctions), in the same house format as regreener/oimpact: a target-first cover leading with a credit-first audit (five things his site gets right, five measured gaps, no viewport meta tag forcing a real 0.398x mobile scale, zero structured data/Open Graph anywhere, a 1.66MB headshot at 185px, no compression/caching, a prose-only corpus with no machine index), a brand-matched concept redesign (JSON-LD per publication/patent, a draft llms.txt, a before/after table) in the same elevated-academic register now shared across all four pages, and, last, short, one paragraph, no ask list, a gentle pitch appendix for EcoWealth's own Ecological Work Protocol (live on Base mainnet 0x76c17C…A14B and Robinhood Chain mainnet 0x5cB9ae…1a9F, a 13,886-def standards-anchored work library, carbon retirement via Klima (hourly until retirements are stopped by operator order as of 2026-08-04), a standard x402 agent-payment economy). The pitch appendix (alexwg-moonshot-brief.html) keeps its full depth, the causal-entropy resonance (his 2013 'causal entropic forces' framework mapped structurally onto verified ecological restoration), plus two more of his works mapped onto the Living Building Challenge and the UN SDGs with live-counted library slices, but is explicitly retitled and framed as the appendix, not the headline. No EcoWealth logo appears anywhere in the package (the round-two base64 starfish brand strip is removed). Every fact about him is sourced from alexwg.org/psi.inc; every fact about EcoWealth is independently checkable (contract addresses, an on-chain paused() check, a live board-stats fetch, per-standard library counts), evidence.json is unchanged from round two.
concept · not affiliated Open the review →Construction & safety industry
A focused wave on construction safety consulting, SSHO staffing, and the profession's credentialing body, passive, evidence-based reviews plus one warm, invited engagement. The through-line: safety documentation produced faster and made tamper-evident, the remote-native lane a certified safety operator can own.
SMART Safety Group
Maintained, error-free site, but invisible to AI buyers (no llms.txt, no business schema) and the living company is trapped in a LinkedIn widget while on-site articles stop at 2022. 4 findings + a credit, honest counter-case included. Warm-in: Sean Puyear.
concept · not affiliated Open the review →Safety Dynamics Group
Genuinely machine-legible for its peer group, yet the careers page renders a broken [awsmjobs] shortcode: an empty supply door for a WBE/DBE firm whose product is certified people. 4 findings + credit.
concept · not affiliated Open the review →EHS Analytical Solutions
SDVOSB / UEI / CAGE / EM 385 credentials live only as prose, invisible to a prime's capture-team AI doing set-aside sub discovery, plus a real elementor JS error on two pages. 5 findings + credit.
concept · not affiliated Open the review →3P Safety Staffing
Already won discovery (real llms.txt, full Organization/EmploymentAgency schema, 200+ city pages), but the 48-hour promise is unverifiable and the intake form omits SSHO and every CSP/CHST lane the site leads with. 4 findings + credit.
concept · not affiliated Open the review →Enviro Safetech
A NASA / Tesla / Stanford roster giving AI zero structured data, no JSON-LD anywhere, llms.txt 404, and money lanes (staffing, consulting) routed to a single mailto. 5 findings + credit.
concept · not affiliated Open the review →BCSP
The profession's certification authority, yet 'is this credential current?' is answerable only through a human form BCSP itself disclaims as non-authoritative; no first-party machine-verifiable credential exists. Respectful opportunity brief, 4 observations.
concept · not affiliated Open the review →SFTCJV, East SF Valley LRT
Warm and invited. LA Metro's 6.7-mile East San Fernando Valley light rail JV, the safety-documentation program shown for the project's safety team: same-day, attributed, hash-sealed daily reports + corrective-action logs, produced remotely.
concept · not affiliated Open the review →CASQA
The QSP credential the construction stormwater world runs on, with no page that assembles the path to earning it. CASQA republished the State Water Board's own 2024 finding that the prerequisites keep new professionals out, yet there is no study guide, only 1 of 21 courses lists a price, and the calendar returns zero events to an AI a newcomer would ask. We built the missing ramp, free to copy. 7 findings.
concept · not affiliated Open the review →California Water Boards
The regulator mandates the credentials and its own 2024 guidance admits the prerequisites block newcomers, but the onramp is missing. The cheapest certified door, a high school diploma at about $120, is buried in an appendix PDF, and the program page names QSD or QSP zero times. We built a six-rung ramp wired to live work, free to copy. Evidence-led brief.
concept · not affiliated Open the review →Digital optimization + dedicated provisioner
Passive, evidence-based reviews of a real business's public surface, the friction a customer or an AI hits, and the fix we'd run as a dedicated always-on provisioner. No EWP lifecycle claims; these are legibility and speed engagements.
Wally's Wine
1.18 MB of JavaScript across 41 files builds the wine catalog in the shopper's browser (stalls on "Loading collection…"); the fix runs on their existing Shopify. The origin engagement.
optimization + provisioner Open the review →Meteora
Michelin live-fire tasting-menu restaurant: the highest-margin private-events lead routes through a shared unbranded form, the phone tel: link is broken by smart-quotes, and there is zero AI-bookability schema on-site. 4 findings.
optimization + provisioner Open the review →ClickSafety
OSHA-authorized training on Adobe Commerce (an Ascend Learning company), /sitemap.xml returns a JPEG, no Course/Offer schema on the flagship OSHA-30 page, llms.txt exists but offers no agent menu (agent-card.json 404); enrollment / DOL-card and 12-seat crew path leaks. 5 findings.
optimization + provisioner Open the review →OSHAcademy
Free-to-study, pay-for-certificate, IACET-accredited with real Course schema on leaf pages, but the only agent-checkout storefront sells merch not certificates, catalog/business pages have zero schema, and the Course schema price ($15.99/$27.99) under-quotes the live checkout ($16.99/$29.99). 4 findings.
optimization + provisioner Open the review →How each one is built, the repeatable method
Passive only, the public site, headers, DNS, agent-discovery probes. No probing, ever.
A handful of findings, each with the exact proof, the redirect, the 404, the DNS record.
A brand-matched demo of the fix, in the org's own elevated look.
The agent-ready files, llms.txt, agent card, MCP tools, they don't have yet.
Every claim re-checked live; every page renders clean; independently audited before it's kept.
Why it compounds
The work is passive and repeatable, we can compress one alongside anything else. Each that lands here is proof, on the record, that the same edge, legibility for humans and agents, with verifiable proof underneath, is missing almost everywhere we look. The portfolio is the argument.
Independent engagement portfolio. Prepared by EcoWealth Corporation on its own initiative. Each review is drawn from the target's own public pages, HTTP responses, DNS records, and published materials, gathered by passive means only, no authentication, no intrusive testing. Each is not affiliated with, endorsed by, or produced in partnership with the organization it looks at; the IRS entry is an independent concept on fabricated data and is not an IRS project, an optimization we attempted unprompted, nothing more. Staged on EcoWealth infrastructure, never deployed to any reviewed organization's systems.