P0: The AI Gateway Building an Agent Economy

P0 is trying to own the layer between people, agents, and AI models.

The easiest way to understand P0 is to start with the gateway. A gateway gives one account and one programming interface access to many AI models.

A developer normally has to open separate accounts with multiple model providers, manage different keys, and rewrite parts of an application when a provider changes. P0's Peezy Gateway offers one OpenAI-compatible endpoint across 29 listed models from 16 providers. It also supports streaming and tool calls, which are needed when a model does more than return text.

P0 then puts products on top of that access. Peezy is a coding agent in the terminal. P0 Editor is a browser workspace. Rapid Mode and the agent API turn ideas into websites or Solana tokens. The terminal, cards, and planned company tools extend the same account into creator and trading workflows.

The best evidence is outside P0's own website.

OpenRouter lists Peezy Gateway as an app with about 611 billion total routed tokens, a number 20 daily rank, and activity dating to December 2025. That does not prove P0 collected revenue on every token. It does prove that the gateway has been used as a real distribution surface.

There is an important limit to that evidence. OpenRouter says app rankings are assigned through an HTTP-Referer header sent with each request. The data confirms that requests reached OpenRouter carrying Peezy's attribution. It does not identify unique people, prove that the official Peezy client generated every request, or show whether the traffic was free, sponsored, or paid.

npm provides a second signal. The public @p0systems/peezy package recorded 38,974 downloads from June 25 through July 24. Its public GitHub repository had 19 stars and eight forks at the cutoff. Those community numbers are small, but the package downloads show that the product is reaching developer environments.

The recent daily series is volatile. P0 routed 133.7 billion tokens on July 17, but only 700.9 million on July 21 and 249.0 million on July 22. Activity recovered to 31.2 billion on July 26 and 16.0 billion on July 27. The correct takeaway is not that usage disappeared. It is that the current base is event-driven and unstable.

Model concentration can be good at the beginning. A gateway often wins by being the easiest place to use one fast, cheap, or scarce model. P0 can then cross-sell other models and tools. The risk is that the most popular supplier changes pricing, opens direct access, or moves elsewhere before P0 has built durable user habits.

P0 is widening the product before the business model is settled.

The team has spent July expanding access, compatibility, and surfaces. That is a distribution strategy: make P0 useful in more places before asking every user to pay.

Investor Log 001 lists an Anthropic-compatible messages endpoint, OpenRouter failover, Kimi K3, xAI Grok, StepFun, and shipped Mac and iOS clients. After that entry, P0 announced free access to Poolside Laguna S 2.1 on July 22, then North Mini Code and Ling 3.0 Flash on July 26. P0's official account has also shown software agents deploying tokens through its API.

The agent stack is broader than a single deployment endpoint. P0's direct documentation shows Rapid Mode for one-call token and landing-page launches, batch deployment for as many as ten projects, creator-fee claiming, hosted .p0.surf pages, and terminal trading. Its published OpenClaw skill lets outside agents register, deploy, claim fees, buy credits, and trade through the same API. This creates a plausible machine-to-machine loop, but P0 does not publish deployment totals, fee income, retained agents, or Pro subscriptions.

The infrastructure underneath this product stack is still rented. Investor Log 001 says every token currently rides rented compute and that P0 was only pricing and siting its first owned GPU nodes. Owned capacity could lower inference costs and create spare-capacity revenue. Until the first nodes are live and their cost curve is published, that is a margin plan rather than an operating asset.

On July 28, P0 announced that it had joined ZaiforStartups. This may improve model access, credits, or partner support, but the announcement did not disclose commercial terms. It is evidence of continued supplier and ecosystem development, not evidence of revenue or financing.

The biggest opportunity is also the biggest evidence gap.

P0 has several ways to earn money. It can charge for model credits, sell Peezy Go subscriptions, charge agent plans, take product or trading fees, and eventually offer team or company products. The public data shows that some billing has started.

The investor endpoint reported $3,938 of usage billed over 30 days, almost all in July. That is early evidence that usage can be priced. The same endpoint reported only nine paid membership orders and $45 collected, however. Billed usage, network value, and collected cash are different numbers and should not be combined.

The most important inconsistency is inside P0's own payload. It labels 48,083 people as paying or active Go members, yet reports nine paid orders. It then displays $480,830 of MRR, exactly 48,083 multiplied by the $10 monthly list price. That MRR should not be treated as realized revenue until P0 reconciles members, orders, collections, refunds, and billing periods.

The newest run-rate claim mixes current products with a future data center.

A July 28 interview clip attributed five new figures to P0's founder: 48,000 active Go members, $480,000 of current MRR, $34,000 of monthly gateway revenue, 2 to 3 billion tokens processed per hour, and a planned data center projected to add $1 million per month.

The arithmetic matters. The two claimed current revenue lanes total $514,000 per month, or $6.17 million annualized. Adding the planned data center produces $1.514 million per month, or $18.17 million annualized. That is how the interview reaches a figure near $20 million. The larger number is a future scenario that includes an unbuilt data center, not a current run rate.

P0's public endpoint adds another reason for caution. It reports about $35,042 of 30-day “network value” but only $3,938 of billed usage. Network value, billed usage, collected cash, and recurring revenue are different measures. The public data does not show which one matches the interview's $34,000 gateway figure.

Credit units also need a definition. Peezy Go advertises 420,000 monthly credits for $10, while the metered gateway lists 10,000 credits per dollar. That would look like $42 if the units were identical, but P0 does not say they are. Sponsored models, unused allowances, or different credit definitions could explain the gap. The report therefore does not infer plan cost or gross margin from credits alone.

This does not erase the bull case. It defines it. P0 has done the difficult first step of finding distribution. If it can show that a meaningful share of those users repeatedly pay, the revenue base could grow quickly from a very small starting point.

P0 reports that liquidity support has started, but investors still need a transaction-level record.

P0's whitepaper says revenue and fees flow into one treasury. At the end of each cycle, operating expenses are subtracted to determine net profit. A set portion of that profit can then fund open-market P0 buybacks or protocol-owned liquidity.

This is a more credible design than a vague promise that “usage helps the token.” It identifies the cash source, the cost layer, and the treasury action. It is also discretionary. The whitepaper does not fix the profit percentage, the buyback cadence, the execution venue, or a public treasury address that reconciles each month.

P0 has moved beyond a roadmap-only claim. Its official communications describe an initial addition of roughly 5 million P0 and 99 SOL, followed by 200 SOL plus roughly 10 million P0 from the treasury. Investor Log 001 calls the larger position protocol-owned liquidity. This is positive first-party evidence. It is not yet independent proof because the report could not match both stated deposits to a published treasury address and exact transaction hashes.

Protocol-owned liquidity is not the same as a buyback. Adding treasury-held P0 and SOL can deepen the market and earn fees, but it does not necessarily purchase P0 from sellers, reduce supply, or give token holders a fixed claim on product revenue. The distinction matters because the current token link is a discretionary treasury policy, not an automatic contract.

The whitepaper also describes future staking funded from monthly profits. That could create a second route from business performance to token demand. Until it is live, it should be treated as a roadmap item rather than current yield.

The investor question is therefore simple: can P0 make the token loop auditable? A monthly statement should show cash collected, model and infrastructure costs, net profit, the chosen allocation, the transaction hashes, and the resulting P0 or liquidity position. One clean reconciliation would be more valuable than another usage headline.

The founder crisis turned the report's liquidity warning into the main event.

P0 traded as high as $0.003797 in the 19:00 UTC candle on July 29. The official account posted its personal insult at 19:11:52. The 19:15 candle closed at $0.002938, down 22.6% from the pre-selloff high. By 21:00, the market had traded as low as $0.000853, a 77.5% drawdown.

The decline continued overnight. The deepest observed low was $0.000352 at 05:00 UTC, 90.7% below the pre-selloff high. P0 later rebounded more than 100% from that low, but the latest sampled close was still 78.3% below the high. A large rebound from a collapsed base is not the same as recovering the original market value.

At the update cutoff, the token was about $0.000854. With roughly 999.96 million units outstanding, that implied an $854,000 token market value. The primary Meteora pool held about $106,000 of liquidity, down 37.9% from the original report cutoff. Twenty-four-hour volume was about $1.58 million, with roughly $845,000 of sells and $738,000 of buys.

GMGN classified the top ten holder rate at 17.62% and the creator balance near 0.31%. The creator-labeled wallet showed no activity after July 18 in the retrieved snapshot. We therefore found no direct creator-wallet sale tied to the incident. That conclusion is narrow: it does not rule out undisclosed affiliated wallets, OTC transfers, or wallets that the data provider did not label.

The liquidity lock still needs a separate caution. P0 says its Meteora LP position is permanently locked. GMGN previously showed a 95% blackhole lock detail while also reporting a zero top-level locked ratio. The incident did not resolve that contradiction. A transparent treasury and LP reconciliation is now more important because investors need to know exactly which controls are independent of the founder.

The product can survive this. The token needs proof that the governance can too.

The asymmetric part of the case is still the gap between visible distribution and a small token valuation. The incident made that gap wider for a reason.

P0 does not need to replace OpenRouter or own a frontier model. It needs a durable niche where users prefer its model lanes, coding tools, local payments, or agent workflows. The investor log emphasizes Southeast Asia, QR and local payment methods, and lower-cost models. That is a sensible wedge because the biggest Western subscription plans are expensive relative to local income.

The product suite can deepen that wedge. A developer can begin with a free model, move into Peezy, store a project in P0 Editor, deploy through Rapid Mode, and later operate as a team. A software agent can use the same system to deploy and earn. Those products can keep compounding even if the token remains impaired.

The token now needs a second bridge: governance. The official account should not operate as an extension of one founder's public behavior. A credible recovery would identify who controls project communications and treasury actions, publish an incident statement, establish review or multisig processes, and show that commercial partners can rely on the company without monitoring the founder's timeline.

What could make the thesis work

• Governance becomes explicit. P0 separates founder, official-account, and treasury controls and publishes a credible incident process.
• Routing becomes habit. Daily usage stabilizes across several models instead of depending on a few large bursts.
• Free users convert. P0 reports repeat paid users and collected cash that reconcile with orders and refunds.
• The suite creates switching costs. Peezy, Editor, Agents, and team tools share projects, history, deployment, and billing.
• Margins improve. P0 moves selected workloads from rented providers onto owned GPU capacity and publishes a lower cost curve.
• The token loop becomes auditable. P0 links each reported liquidity addition or buyback to its treasury address, transaction hashes, and monthly profit calculation.

Why the starting valuation matters

An $854,000 token does not require enterprise-scale profit for cash flow to become material. A hypothetical $25,000 monthly profit allocation would equal about 2.9% of the update-cutoff market value before price impact. This is an illustration, not a forecast. P0 has not disclosed a fixed allocation, proven that profit level, or shown that token support would take priority after the incident.

The same small size cuts both ways. It creates upside if governance and operating evidence improve, and it allows another confidence shock to overwhelm the pool.

The thesis now breaks on governance before it breaks on economics.

• Founder and communications control. The official account amplified the founder dispute instead of containing it. Without separate controls and a crisis process, one person can damage the product, partner relationships, and token again.
• Partner risk. P0 depends on model suppliers, infrastructure providers, developers, and the Solana ecosystem. Public hostility can make counterparties less willing to support or integrate the project.
• Metric credibility. The paying-member, order, cash, and MRR fields do not reconcile. Repeated inconsistencies would make every first-party operating claim less useful.
• Free or subsidized demand. Large routed-token totals can come from promotions, cheap models, or subsidized traffic. Volume without positive unit economics can destroy value.
• Supplier concentration. One model accounts for 47.4% of routed tokens. A pricing, capacity, or distribution change could reduce the gateway's strongest lane.
• Gateway competition. OpenRouter, direct model APIs, coding agents, and cloud platforms can copy compatibility features. P0 needs workflow depth or regional advantage.
• Product sprawl. Coding, token deployment, trading, cards, and company tools can reinforce one account, or they can divide a small team across too many products.
• Rented-compute dependence. P0 currently depends on outside inference capacity and failover providers. Pricing, access, or reliability can change before owned infrastructure exists.
• Discretionary token economics. Reported liquidity additions are constructive, but a “set portion” of profit is not a fixed token claim. Management can change the allocation, delay it, or prioritize operating needs.
• Thin liquidity. The primary pool fell to roughly $106,000 of liquidity. The July 29 selloff showed how quickly entries and exits can move price during stress.
• Execution and uptime. P0 has publicly discussed downtime during demand spikes. Reliability is part of the product when developers depend on one endpoint.

Watch governance first, then return to conversion.

The product kept running. The founder and official account created a governance crisis.

This was not a disclosed exploit. It was a public-control failure that became a market event.

The sequence began with the founder account promoting P0's OpenRouter ranking and comparing its routed-token volume with much larger software companies. After community members repeatedly tagged Solana ecosystem builder Vibhu, he said he was happy to help but that a relationship takes two. P0's official account then replied with a direct personal insult at 19:11 UTC.

The official account did not quickly separate the project from the founder's behavior. At 19:53 UTC it described the founder as vocal while calling P0 the largest AI project on Solana. Ten minutes later it framed the dispute as other people simply not liking P0. At 02:11 UTC on July 30, the founder account offered only a partial apology and continued insulting the Solana team.

A public collaborator later defended the founder's work ethic and said the project would keep moving. That supports continuity, not resolution. We found no public announcement of a resignation, account-control change, independent board action, treasury action, product shutdown, or postmortem at the evidence cutoff.

The report separates what is visible from what is promised.

Methodology and limitations

The report used the exact P0 mint and primary pair supplied by the user. Other crypto and software projects also use the name P0. They were excluded unless P0 Systems linked to them directly.

OpenRouter independently displays the aggregate routed-token count, app rank, active-since date, model count, and category rankings. P0's public stats endpoint provides the daily series and mirrors the OpenRouter aggregate. OpenRouter attributes requests through application-supplied headers. It does not verify unique users, P0 revenue, paying members, or owned compute.

P0's operating, membership, billing, and signup fields are first-party public data. The report preserves the contradictions instead of choosing the most favorable field. No private billing system, bank statement, or audited financial statement was available.

The July 28 interview figures are founder-attributed, first-party claims reported by BNN. The report reproduces their arithmetic to separate current product claims from a future data-center forecast. It does not treat the interview as audited financial evidence.

Investor Log 001 and official P0 posts are first-party release records. They support that announcements were made, including reported protocol-owned liquidity and new model access. They do not independently verify revenue, program terms, treasury ownership, or transaction-level execution.

The founder-incident timeline uses direct posts from the founder account, P0's official account, Vibhu, and a public collaborator. Grok Heavy was used only to discover leads. Every incident claim included in the article was checked against the direct post. The report describes observable conduct and does not diagnose the founder's mental state.

GMGN provides exact-pair 15-minute candles, liquidity, holder classifications, creator-wallet activity, and security fields. Its wallet labels and risk tags are algorithmic. The timing supports a confidence-shock interpretation, but it does not prove that social posts were the only cause of every trade. No creator-wallet activity was found after July 18 in the retrieved snapshot; that does not rule out undisclosed affiliated wallets.

The $854,000 token value multiplies the observed update-cutoff price by the 999.96 million reported supply. It is not enterprise value, and the P0 token does not represent a legal equity claim on p0 Systems.

This report is for information and research. It is not investment advice or an offer to buy or sell an asset. P0 is an early, thinly traded token. Product status, market data, and token mechanics can change quickly.