---
name: zzz-xvix-gmx-dev
description: Proof that the ZZZ (Robinhood Chain) deployer is xvix.eth, the wallet that deployed the GMX token, plus the corrected DD. Read before any ZZZ, Autonomous Protocols, protocol.link, @ExponentLabs_, @auto_protocols or GMX-founder question, and before any "is the dev a fresh wallet" call on a new chain.
sources: [cowork]
aliases: [ZZZ, zzz protocol, Autonomous Protocols, protocol.link, ExponentLabs, xvix, xvix.eth, GMX dev, auto_protocols]
---

## ⚠ CORRECTION to the ZZZ row in [[screen-log-8]], same day, 17 Sep 2026

I wrote "**nonce 0 at launch, no prior launch exists from this address**, so the rumour is not supported by anything on chain". **That was wrong, and the number that misled me was the nonce read on the WRONG CHAIN.** `eth_getTransactionCount` on Robinhood Chain was 0 because the chain is 70 days old, not because the wallet is new.

**STANDING RULE, and it is mechanical: on a NEW chain, a deployer's nonce means nothing. Before calling any wallet fresh, read its nonce, balance and ENS on Ethereum, Base, Arbitrum and Optimism as well.** One batched `eth_getTransactionCount` across four public RPCs, about five seconds, and it is now a required step in [[dd-process]] question 1. Second cheap step that settles identity outright: **ENS reverse lookup** (namehash of `<addr-without-0x>.addr.reverse`, `resolver(bytes32)` on the ENS registry `0x00000000000C2E074eC69A0dFb2997BA6C7d2e1e`, then `name(bytes32)` on the resolver). `eth.blockscout.com/api/v2/addresses/{a}` also returns `ens_domain_name` directly and confirmed it independently.

## The proof: the ZZZ deployer deployed GMX

Wallet **`0x5F799f365Fa8A2B60ac0429C48B153cA5a6f0Cf8`**, ENS **xvix.eth**.

- Nonces: Robinhood Chain 0, **Ethereum 826, Arbitrum 10,862**, Base 19, Optimism 2.
- **It is the `creator_address_hash` of the GMX token on Arbitrum `0xfc5A1A6EB076a2C7aD06eD22C90d7E710E35ad0a`.** Not a contributor wallet, the deployer.
- Also deployed on Ethereum: **XVIX `0x4bAE380B5D762D543d426331b8437926443ae9ec`**, **XVIX GMX (IOU) `0x491050F83F54FF86907E77732Da7964E870fE1C1`**, **3X ETH/USD X2 FS `0x03B8F104A9bcf1cEb1140F20ef0f43Ccc787d8b0`**, **X2Fee `0x2f14b124C6957F2199039b60ca2c2eBe501Ef4FC`**. That is the whole XVIX → X2 → Gambit → GMX lineage from one wallet.
- Holds **26,621.8204 esGMX** (`0xf42Ae1D54fd613C9bb14810b0588FaAa09a426cA`, = $193,274 at GMX $7.26) and 26.0816 GMX on Arbitrum; and on Ethereum the migration IOUs **UNI GMX 344.4535** (`0x2eF13496946e9a2150977C855F9E34b5aB945b31`) and **XLGE GMX 123.75** (`0x3c6930d6A4D698B9f4C24b4E14d0f9CF9DaC9c68`), plus 90.2786 ENS.
- On X the ZZZ community addresses @ExponentLabs_ directly as "**Hey X**", the known pseudonym. (@xdev_10 is an empty 102-follower account, not him.)
- **GMX today: $75,976,981 market cap, $7.26.** It was one of the very few DeFi products with large sustained real revenue and genuine fee-sharing (30% of platform fees to GMX stakers). So "a dev who ran a huge project before" is TRUE and understated.

**So Ben's rumour was right and my first pass was wrong.** Everything else in the [[screen-log-8]] ZZZ row (gate clean, 11,858 holders, $471,251 of fees, the 170M burn, the fee-recipient move today) still stands unchanged.

## What this changes, and what it does not

**Changes:** the operator grade goes from "anon with no history" to the single most credentialed builder screened on this chain. His whole track record is a token that pays holders a share of real fees, which makes the ZZZ FAQ's "proposed" revenue-to-buyback far more credible than the same sentence from an anon. He also burned 170,000,000 (20.24%) of his own 227,586,207 allocation and kept only 6.86%.

**Does NOT change:** he has taken **194.1034 ETH = $471,251** of Pons creator fees in 12 days and returned **zero** to holders, and on 17 Sep 08:38 BKK he moved the fee recipient to `0x698084856F65Bf1B4740E3D0dBA0B57A5B1deBA9`, a wallet that is **nonce 0 on Ethereum, Base, Arbitrum and Optimism** (so brand new, could be a personal cold wallet or a new project treasury; unresolved) and which already holds the 56,000,000 ZZZ and 186.1 ETH he sent it.

**New risk his own history creates: he MIGRATES tokens.** XVIX → Gambit → GMX, with IOU claim tokens for each hop. A ZZZ holder should expect a migration to be on the table rather than assume this ticker is the terminal one.

## The project

- **zzz.protocol.link** is live: a token-gated community, minimum **4,880.43 ZZZ = "$100 at today's membership price"**, recalculated daily and frozen per member at join, tokens stay in the holder's wallet, no deposit. Holdings passively generate credits, 1 credit per message.
- **protocol.link** = "Autonomous Protocols, the coordination platform for people and AIs": discussion → AI-generated tasks → tasks completed by members or AIs. Explicitly aimed at replacing the Telegram → forum → Snapshot → Tally loop DAOs use.
- Accounts: **@ExponentLabs_** (Oct 2020, 14.1K followers, 99 posts, bio "Supporter of ZZZ") and **@auto_protocols** (Sep 2026, 735 followers, 7 posts, "Created by the ZZZ community"). Shipping is visible and dated: 13 Sep answering bot on the FAQ, 16 Sep improved automated task generation.
- Token is the stock **PonsV2LauncherToken** template, zero custom contract work. All the work is off-chain product.

## Verdict after the correction

**Small speculative buy on the operator, not on the numbers.** $14.7M ex-burn cap (DexScreener's $18.2M counts 170M burned tokens still sitting at the zero address), -77.5% off a $65.6M ATH set 12 Sep, sellers leading buyers 2,036 to 1,489.

**The one thing to watch, one address: a ZZZ buy leg appearing from `0x698084856F65Bf1B4740E3D0dBA0B57A5B1deBA9`.** That converts $471k of proven revenue from his income into the token's value accrual and is the whole thesis. Absence of it three weeks from now is the falsifier.

**Source that failed: twitterapi.io returned HTTP 402 Payment Required all session**, so every X read here was done by hand through the Chrome bridge and the coverage is not exhaustive. Ansem was checked and has NOT posted about ZZZ.

## What the FUD actually was, and it was NOT the market. Reconstructed 17 Sep 2026 once twitterapi.io credits came back.

**⚠ CORRECTION to my own line in this file: I wrote "Ansem was checked and has NOT posted about ZZZ." Wrong method, wrong conclusion. He never POSTED it, he said it on Market Bubble ep.19 and BOUGHT $50,000.** Quoted by @Kaiz_294 12 Sep (312 likes, 225,550 views) and @ZygfridS (50,483 followers): *"I did drop that into $ZZZ, X dev is like one of the core devs behind GMX."* **STANDING: `from:handle` only catches what a KOL typed. A podcast or Spaces call is invisible to it. Search the TICKER plus his NAME as well, always.**

**THE MARKET DID NOT DO THIS. Daily closes 12 Sep to 17 Sep: ZZZ -60.9%, PONS +4.9%, AI -4.7%, PRM +31.3%, HOOKR -18.2%, BTC -1.0%, ETH -3.2%, SOL -2.6%.** Same chain, same days, everything else flat. ZZZ fell alone. (From the 12 Sep intraday high $0.09798 the drawdown is -77.5%.)

**THE CAUSE WAS A PRODUCT REVEAL THAT DISAPPOINTED, ON THE EXACT DAY OF THE TOP.**
- **12 Sep 00:24 UTC** @ExponentLabs_ posts the first look at the platform, 459 likes, "There will be a longer description soon". Price is at ~$70M, within 11% of the $78.9M ATH.
- **12 Sep 00:33** @ScruFFuR (56,311): *"the average $zzz holder will see this and think 'that's it?'"* **12 Sep 00:47** @latentliq: *"At first glance it looks like a token-gated forum/governance platform, but I don't think this is the actual AI product."* The doubt starts within 25 minutes of the reveal.
- **12 Sep 05:31** the Ansem $50K buy goes viral. **That is the top tick.**
- **12 Sep 13:04** @Donaxbt (39,564) opens the bundler strand: *"average entry point for $ZZZ bundlers is $3m market cap, realized PNLs from $100k to $475k and remaining holdings of $100k minimum waiting to be dumped."* Note this is a $3M-cap cohort, NOT a launch bundle, and it does not contradict the clean 15-second snipe gate.
- **12 Sep 14:02 and 15:27** @nastyboyyzz: *"$zzz fud happening"*.
- **12 Sep 19:50** @DjGriffith (10,671) from the rival `0xf44702b17d9abd53815f703e772f35e9c71a53af` camp (Ethereum Foundation backing, ex-Lido devs): *"$ZZZ is just a meme with no product tied to it."*
- **13 Sep 04:40** ExponentLabs_ posts the full 16-part product thread. **Engagement collapses against the teaser: 161, 45, 28, 17, 16, 15, 15, 14, 14, 13 likes versus 459 the day before.** The detailed explanation got a fraction of the attention of the tease.
- **13 Sep 09:05** @heeshilio: *"Is it just me or is the $zzz product announcement pretty underwhelming?"* **13 Sep 12:05** @dragonsdennnn sells out at a $15M average: *"Thought the project would be more than just a group chat that Claude can make in 30 seconds."*

**So the FUD had three strands and only one of them was substance:**
1. **DAY ONE, RESOLVED: "is this really X or a hacker holding the GMX deployer key?"** @sardonic_owl 5 Sep 00:45 also found a real, concrete problem: **ZZZ's then-listed website served a fake Cloudflare check telling visitors to press Windows+R and paste a PowerShell command** — an actual malware page, not a rug claim. @latentliq 5 Sep 10:06 reported that an ex-GMX dev confirmed via Axe that ZZZ is genuine and not hacked. Settled, and my own on-chain work independently proves the wallet deployed GMX, XVIX and the migration IOUs.
2. **11 Sep, REAL AND STILL OPEN: the deployer moving money.** @johnyjazz123 (549 followers, 10,728 views) flagged exactly what I found: *"xvix.eth sent 0x698084856F65Bf1B4740E3D0dBA0B57A5B1deBA9 the ZZZ deployer fees, 100 ETH, and 56M ZZZ. What is Xdev planning?"* Nobody large picked it up.
3. **12-13 Sep, THE ONE THAT DID THE DAMAGE, and it is not a scam claim at all: the product is a forum.** A $82M cap had priced "GMX founder ships an autonomous AI protocol"; what shipped was a token-gated discussion board with AI-generated tasks.

**✅ ONE BULLISH MECHANIC THE FUD BURIED, from the 13 Sep thread: "Bonus points, pioneered at @GMX_IO. Token holdings passively generate bonus points every second... Bonus points are burnt on sale of tokens proportionally."** That is GMX's Multiplier Points re-implemented: a real, tested, pro-holder anti-sell mechanic from the one founder who has shipped it before. It is the strongest argument that the "proposed" revenue-to-buyback is not just a placeholder.

**Verdict unchanged from the corrected DD: small speculative buy on the operator.** The drop was a reveal that undershot expectations, not a scandal and not the market. The falsifier is still the single address `0x698084856F65Bf1B4740E3D0dBA0B57A5B1deBA9` and whether ZZZ ever gets bought from it.

## Reality check, 17 Sep 2026. Ben: "separate the emotions from the reality."

**✅ THE OFFICIAL GMX STATEMENT SETTLES THE KEY-THEFT QUESTION. @GMX_IO, 8 Sep 2026, 689 likes, 266,415 views, `https://x.com/GMX_IO/status/2097348449914024279`.** Full text, verbatim: *"X, GMX's original lead developer, has stepped back from active development and is nowadays an advisor to GMX. **He has personally confirmed that he deployed $ZZZ on Robinhood Chain this weekend.** The wallet used for that is not an active GMX deployer and is not connected to any active GMX V2 contract. It is an original personal deployer, the same one also previously used to launch XVIX, a project that preceded GMX. GMX moved to fresh deployer wallets long before this deployment, as part of an OpSec upgrade implemented in coordination with X. Those new wallets are not controlled by individuals and have additional security safeguards. **X has renamed his personal Twitter account to @ExponentLabs_** and has started setting out an AI-focused vision for autonomously evolving protocols there. He is building this project **outside GM Labs (@GMLabsHQ)**, the company behind GMX and GMTrade, so **GMX has no link to it beyond his advisory role.**"*
Corroboration: the wallet was never swept (still holds 90.2786 ENS, 4,647.9 ARB, 26.08 GMX, 576.6 USDC.E, 51.55 LINK) — a thief drains, he does not sit for 12 days. And @ExponentLabs_ was formerly **@xdev_10**, now an empty 102-follower husk, exactly matching "renamed his personal Twitter account". **Probability of a stolen key or hijacked account: effectively zero.** The cost of the statement is the other half: he is a SEMI-RETIRED advisor doing a side project with zero GMX resources behind it.

**✅ THE "SCAM PAGE" DOES NOT SURVIVE CONTACT. Single source: @sardonic_owl, 24 followers, 3 likes, 5 Sep 00:45, and the post does not even contain the domain** — it literally reads *"ZZZ's listed website, , serves a fake Cloudflare verification"*, with the URL missing. At 7 hours old @0xFomoNaut recorded *"No website, no X account, no thread explaining itself"*, and DexScreener's `websites` array for the token is **empty to this day**. So there was no official site for a malicious page to be. Live check today: `zzz.protocol.link` and `protocol.link` both return 200 with a **hand-written strict CSP** (`script-src 'self'; object-src 'none'; frame-ancestors 'none'; base-uri 'self'; form-action 'self'`) and no interstitial of any kind. **Verdict: not his, never corroborated, and the tightest CSP I have seen on a Robinhood-Chain project is the opposite of a drainer page.**

**⚠⚠ THE BIGGEST FINDING: HE UNDER-PROMISED AND THE MARKET OVER-PRICED. This is the whole story of the -77.5%.** On day one, **5 Sep 23:22 UTC, 289 likes, 102,797 views**, he wrote: *"A few years ago, had a vision for autonomously evolving protocols. Seems we now have the technology to build it. **First simple part probably takes a week or so.**"* He shipped the first simple part on 12 Sep, **exactly one week later, exactly as described**. The market had bid $82M on the phrase "autonomously evolving protocols" and then punished him for delivering the thing he actually said he would deliver. **There is no broken promise anywhere in the timeline. STANDING: when a token collapses on a "failed" delivery, pull the founder's FIRST post about it and check what was actually promised before accepting the crowd's framing.**

**✅ "JUST A GROUP CHAT CLAUDE COULD MAKE IN 30 SECONDS" IS FALSE, MEASURED.** App bundle `zzz.protocol.link/assets/index-rR3O0w6c.js` is **494,682 bytes**. Separate backend at **api.protocol.link**, `/health` → `{"ok":true}`, everything else namespaced behind auth, plus a realtime **WebSocket** (`wss://api.protocol.link`, allowed explicitly in the CSP `connect-src`). Route strings pulled from the bundle: `/auth/challenge`, `/auth/verify`, `/auth/logout`, **`/auth/passkey/options`, `/auth/passkey/verify`** (42 passkey / 4 webauthn references), `/membership`, `/posts`, `/post/`, `/vote`, `/tasks`, `/task/new`, `/task-invitations`, `/agents`, `/api/realtime`, `/api/files/`, `/api/profiles/`, `/api/interest`, `/admin/settings`, `/admin/moderators`, `/admin/public-faq`, **`/admin/ai-instructions`**, and **`/admin/task-lab`, `/admin/task-lab/discussions/`, `/admin/task-lab/runs`, `/admin/task-lab/snapshots`**. Keyword counts: task 318, role 94, credit 56, claim 51, vote 47, moderation 13, bonus 12. **The task-lab with RUNS and SNAPSHOTS is the tell: that is an eval harness for tuning the discussion-to-task model. Nobody builds run-and-snapshot tooling for a chat room.** HONEST OTHER HALF: what a visitor actually SEES is a forum, because all the depth sits in the pipeline and the admin surface. The complaint is fair about the surface and wrong about the build.

**THE VISION, IN HIS OWN WORDS, ALL SOURCED.**
- **15 Sep:** *"What's been built: Reddit but more productive. **The end goal: Cursor for communities.**"*
- **13 Sep thread:** discussion → AI turns it into tasks → tasks auto-completed by AIs or claimed by members individually or in teams → contributions recorded and rewarded. *"Token communities are a natural place to start... at the same time, we will work towards **abstracting away the blockchain layer to enable any community, organisation, or team** to use the platform."* And: *"much of the long-term value will come from the quality of the AI integration: how much of project management and task execution can we automate, and how precisely?"* So the target market is explicitly not crypto.
- **9 Sep, 180 likes, the ideological core:** *"(More capital / power) → greater ability to generate returns → (More capital / power)... In earlier times, buying a factory / real estate. In more recent times, access to private markets. **Now, access to large swarms of agents.** A way to bring some balance to this would..."* He is trying to give ordinary groups access to agent swarms.
- **TOKEN SINK, stated plainly 13 Sep, 177 likes:** *"For any questions about ZZZ token utility: **Platform subscriptions → ZZZ buybacks**."* Extended 15 Sep when he endorsed a member's proposal: subscriptions buy ZZZ **and** AI API credits; communities **stake ZZZ to generate credits**; credits pay for AI usage. That makes ZZZ a prepaid meter for compute, which is a real sink rather than a governance fig leaf.
- **Anti-sell mechanic:** *"Bonus points, pioneered at @GMX_IO. Token holdings passively generate bonus points every second... burnt on sale of tokens proportional to the amount sold."* That is GMX Multiplier Points, already battle-tested by him.

**⚠ THE HONEST WARNING, and it comes from his own record. The "vision from a few years ago" is X4, and X4 was never shipped.** Independently confirmed from 2022: DefiLlama's round-up, 9 Apr 2022, *"GMX X4: A Protocol Controlled Exchange"*; @TheDeFinvestor (166,054) wrote a thread on it 25 Jul 2022; @PrismaticCap called it a revolution of the AMM model. @NintendoDoomed, Mar 2023: GMX pivoted to synthetics and **X4 was shelved**, with LionDEX shipping the PvP-AMM concept first. So the pattern is real: he proposes large architecture, and at least once he did not finish it.

**⚠ AND THE HARD GAP: NOTHING IS ON-CHAIN. There is no staking contract, no credits contract, no buyback contract, no protocol.** His wallet has 18 transactions on Robinhood Chain and has deployed nothing but the token through the Pons factory. "Autonomously evolving protocols" is today a web2 application with a token gate in front of it, while **$471,251 of creator fees have gone to him and $0 of the stated buyback has executed.**

**NET: the delivery was honest, the product is well built, the drop was a pricing error, and the remaining risk is entirely execution and value-accrual, not fraud.**

## Benchmarking pass, 17 Sep 2026. Ben: competitors, ease of copy, moat, dev differentiation, does the token help, what if Microsoft builds it.

**CATEGORY: AI coordination / agent tooling with a token-gated membership wrapper.** Not a launchpad, not DeFi. Return bucket: holder must operate nothing, but nothing pays automatically either, so it sits in the weakest accrual band until the buyback is code.

**RING 1, same chain: EMPTY.** No other Robinhood Chain project does discussion-to-tasks. He owns the slot outright, and on this chain that has been worth something.

**RING 2, same job, OFF chain, and this is the problem: THE INCUMBENTS HAVE ALREADY SHIPPED IT.**
- **Microsoft: `Planner Agent` inside Microsoft 365 Copilot; `Facilitator` in Teams meetings auto-capturing action items; Copilot turning meeting transcripts into Planner tasks.** Verified 17 Sep 2026 via techcommunity.microsoft.com and support.microsoft.com.
- **Slack / Salesforce: Slackbot rebuilt as "the ultimate AI teammate", plus Agentforce agents running as teammates inside Slack channels.**
- Also Notion, Linear, Asana, ClickUp all shipping AI task extraction.
**So the answer to "what if Microsoft or Google built this" is: they already did, and it is bundled free inside a seat the customer already pays for.**

**RING 2, same job, CRYPTO side, and the comp is brutal: Olas / Autonolas ("Co-own AI", the decentralised AI-agent coordination protocol, 18,754,350 agent transactions) has a token market cap of $8,121,392 — BELOW ZZZ's $14.7M ex-burn**, after years of building and real usage. Dework and CharmVerse (token-gated DAO task managers) still exist and are tiny. **The crypto version of this exact idea has been tried repeatedly and never got big.**

**RING 2 ceiling, AI-agents category (CoinGecko, 17 Sep 2026): VVV Venice $1,194,515,356 · VIRTUAL $394,533,274 · FET $362,733,713 · KITE $254,144,767 · ARC $72,982,034 · HOLO $43,836,929 · OLAS $8,121,392.** ZZZ at $14.7M sits in the lower third of a category whose top is a billion.

**RENTED vs OWNED.** RENTED: the entire AI layer (he buys model API credits — his own 15 Sep post proposes subscriptions funding "AI API credits"), plus hosting and Cloudflare. OWNED: the coordination product itself (forum, task pipeline, the `/admin/task-lab` eval harness with runs and snapshots, passkey auth, membership logic that freezes each member's threshold), the ZZZ lore, the 11,858 holders, and the founder. **The OWNED column is real but it is a product, not a protocol: there is not one contract deployed.**

**EASE OF COPY: a competent team rebuilds the VISIBLE product in 2 to 4 weeks with AI help.** The genuinely hard part is not code: it is making the discussion-to-task output good enough that people accept the tasks instead of ignoring them. That is what the task-lab runs and snapshots exist for, and it is months of iteration. **But Microsoft and Slack have exactly that tuning capability at a thousand times the scale and a thousand times the training data.**

**MOAT, stated separately as required.**
- **Product moat: weak.** No data nobody else has, no integration lock-in, no named integrator, no switching cost, no licence. Being first on a small chain is not a moat.
- **Token moat: currently ZERO.** The fee link is a sentence on a website ("Platform subscriptions → ZZZ buybacks"), not a take rate readable on chain. The toll booth is a plan.
- The only durable asset is the operator plus the community.

**⚠ THE VENICE / $VVV "COLLABORATION" IS UNCONFIRMED AND IS BEING TRADED AS IF IT WERE NEWS.** @AskVenice has said **nothing** (`from:AskVenice (ZZZ OR ExponentLabs)` returns 0). The source is @yh_0x on 14 Sep, 122 likes, whose own wording is *"after collaboration ... **is confirmed**"* — future tense. Treat as rumour until Venice posts.

**⚠ BUILDER SIGNAL IS ESSENTIALLY ZERO. Of 80 replies to @ExponentLabs_ and @auto_protocols since 10 Sep, 2 were technical and 78 were trader noise.** No API questions, no integration requests, no self-hosting questions, no bug reports with use cases. The audience is a trading audience, not a user base.

**PMF VERDICT, split because it genuinely splits:**
- **For companies: DEMAND EXISTS, ALREADY CAPTURED.** Every named buyer already has Copilot or Slack AI bundled into a seat they pay for, and no company puts internal planning behind a volatile memecoin. Procurement alone kills it.
- **For ownerless internet groups (token communities, open-source collectives, friend groups, DAOs): DEMAND EXISTS, UNSERVED.** These have no employer, no HR, no procurement, no Microsoft tenant, and no incumbent serves them. That is the wedge, and it is the one he actually chose.

**DOES THE TOKEN HELP OR HURT? Both, and the split is clean.**
HELPS: it funded the whole thing instantly and without dilution ($471,251 in 12 days, no VC, no board); it delivered 11,858 users on day one who are financially motivated to use and market it; token-gating is a genuinely good anti-spam membership primitive; and staking ZZZ to mint AI credits is a real consumption sink, not a governance fig leaf.
HURTS: it caps the customer at "people who will hold a memecoin to enter a chat"; it brought traders instead of builders (2 of 80); it puts a semi-retired founder on a price clock, which is how large architectural ideas get rushed; and **his own roadmap says he will "abstract away the blockchain layer" for normal teams, which means the token is the bootstrap and can be left behind by the very success it is funding.**

**WHAT IS ACTUALLY BEING BOUGHT: an operator and a narrative slot, not cash flows.** Size it as that.

## Ben's correction, 17 Sep 2026: "it's not the AI making task items, it's that the AI actually gets tasks done." He is reading the vision right, and it changes the competitive frame.

**He is right about the CLAIM. Part 3 of the 13 Sep thread: *"Automatic task completion. Some tasks can be automatically completed by AIs. But some tasks still require a human touch, e.g. making an introduction or connection. Tasks can be opened for claiming."* And the stated end state, 15 Sep: *"What's been built: Reddit but more productive. **The end goal: Cursor for communities.**"* Cursor is an execution analogy, not a note-taking one. So execution IS the pitch.**

**But he himself separates built from goal, and everything SHIPPED so far is the creation half.** His dated shipping log: 13 Sep answering bot, 15 Sep passkey logins, **16 Sep "The automated TASK CREATION system has been improved — task descriptions are now more precisely generated based on discussions."** Not one shipping post is about execution, and there is no public demo anywhere of an AI completing a task end to end.

**What the client bundle shows about the execution loop:** `autoComplete` 10, `review` 53, `output` 39, `approve` 15, `reject` 11, `instructions` 15, plus the `/admin/task-lab` runs and snapshots. **So the implemented shape is: AI drafts an output, a human approves or rejects it.** Absent from the client: MCP, tool_call, sandbox, repo or pull-request machinery, browser automation, any model-provider name. **Caveat stated honestly: execution would run server-side at api.protocol.link, so client-side absence is weak evidence, not proof.**

**WHY THE DISTINCTION MATTERS, and it changes the Ring 2 answer I gave earlier.**
- If the product is task CREATION, Microsoft Planner Agent and Slack/Agentforce win trivially and it is over.
- If the product is genuinely AI EXECUTION, the peer set becomes Devin/Cognition, Claude Code, OpenAI agents, Manus, n8n/Zapier agents, **and Agentforce, which already takes actions inside Salesforce** — so the incumbents do not vacate the field, they just fight on different ground.
- **The one thing incumbents structurally cannot do: they execute only inside a tenant's own walled systems (your CRM, your files, your calendar). A group of strangers on the internet has no shared company system to execute against. That gap is real and unserved.**
- **The counter-risk, and it is the honest one: the actual tasks of an internet community are "make an intro", "design a logo", "write the thread", "reach out to X". Those are either human-only or generic-LLM work that anyone can do in ChatGPT for free. The value of automating them may be much smaller than the phrase "AI does the work" implies.**

**So the thesis in its strongest honest form: if he cracks group-discussion → agent-swarm-execution, with ZZZ staked as the metered credit for compute, the token is the meter on a real business. Today that is a stated end goal with the creation half shipped and the execution half unproven. That is exactly the gap being priced, and it is why this is an operator bet rather than a cash-flow bet.**