Last week the US government threatened to destroy an AI company for refusing to enable mass surveillance. This week I'm writing about the company that's architecturally immune to it. A $VVV / @Venice_AI investment thesis 🧵

I think @AskVenice “Minds” feels like the right direction after this Fable debacle. Open source doesn’t win by producing one model that cleanly beats Claude/GPT. It wins through systems: open models, routing, specialized agents, evals, verification, and private inference. Cheaper models for simple tasks, stronger reasoning when needed. Agentic chat is where users first feel that shift and it’s impressive.

Space agriculture just got sweeter. 🍓 Redwire has been awarded a contract from Astrobiome Space to launch space agriculture mission in the Redwire Greenhouse, the World’s first commercial space greenhouse. Astrobiome Space will use Greenhouse to grow the first wild-strawberries ever cultivated in orbit on the @Space_Station. This will be the first flight for Redwire’s Greenhouse. “This contract with Astrobiome Space marks an exciting step forward in advancing critical technology for sustainable life-support systems beyond Earth,” said Marc Dielissen, Executive Vice-President of Redwire Europe. “With our Greenhouse platform, we are not only enabling fundamental research, but we are also demonstrating the practical cultivation of fresh food in space—including crops such as strawberries and fungi. These capabilities are essential as we prepare for long-duration missions and future habitats, while also generating valuable insights to improve agricultural practices here on Earth.” Read more in the press release: bit.ly/4vwCmgT

There are much more interesting asset classes now than crypto. Find the ones that are just starting their bubble phase. Early crypto vibes.

What a post! All of Europe should take note lest we all become/continue to be “museum hosts”.

venice ($VVV) just had an insane run-up. fair question: is this overvalued? let me actually do the math the chart shows venice's daily text inference over time. january started at ~10B tokens/day. today it's ~55-65B/day — a 6x increase in 4.5 months. that's the headline. now let's figure out what it means for the token. the attached table breaks that same metric down by costcode (each costcode = one model). the AVG column is average tokens/day per model. summed across all 29 visible costcodes you get ~55B/day, which roughly matches the recent bar heights on the chart — so we are using this as a proxy to get the model mix for the current usage from there the math is straightforward: - map each costcode to its venice catalog price (routed variants like parasail/deepinfra/openrouter all price at the base model rate) - blended price = 75% input 25% output (typical chat/agent mix) - AVG × 365 × blended price, summed across all models = ~$52M annualized but $52M isn't venice's revenue. it's the API-equivalent VALUE of all that text inference if every token was billed at catalog rates. and the reality is more nuanced because venice has multiple ways usage gets paid for: - Pro ($18/mo): unlimited access to base/open-source models (GLM, Qwen, Mistral) ~$10/mo of frontier model credits - Pro ($68/mo): ~$75/mo of frontier credits - Max ($200/mo): ~$2,25/mo of frontier credits - DIEM holders: perpetual daily credits when locking up VVV supply, $0 cash to venice when consumed - direct API customers: pay catalog per-token this is actually clean unit economics. the free unlimited stuff is on cheap-to-serve open-source models. Claude Opus, GPT-5, etc are metered — either via direct API, Pro /Max credits, or VVV staking. so the expensive inference on the chart (Claude alone = ~$31M of the $52M ceiling) is mostly captured as revenue. the cheap open-source inference is what gets subsidized via Pro on-chain sanity check: buy-and-burn ran ~162K VVV in 4 months ≈ $2.3M of revenue burned ≈ $7M annualized burns. if burns are 15-30% of revenue, total cash run rate is ~$25-50M. matches voorhees' february commentary (<$48M ARR) plus growth since put that against the valuation: - $650M MC = 13-26x revenue - $1.13B FDV = 23-45x revenue now compare to the private inference platform peers: - openrouter: 26x ($50M ARR raising at $1.3B) - fireworks: 31x ($130M ARR at $4B) - together AI: ~25x ($130M revenue at $3.3B) - AI-native private median: 28x snapshot verdict: venice sits right in the peer band. after a major rally, that's fair value. not cheap, not overpriced. so on multiple alone the run-up looks earned but not deeply mispriced anymore but the peer comparison falls apart the moment you look at what venice is and what it isn't: MINIMAL DISTRIBUTION. no enterprise tier, no team plan, no big sales motion. openrouter, fireworks, and together all run real enterprise sales adding 50% to ARR. venice is pure self-serve crypto-native / organic distribution. the easy revenue lever literally hasn't been pulled. and they just launched Pro and Max tiers in april — the upsell motion is barely started PRIVACY IS EARLY. with recent breaches and mass usage of AI, the privacy story is only going to get stronger over time. venice is the only architecturally-private inference layer and consumer app on the market — not privacy as policy, privacy as math (TEE E2EE). that moat compounds with every breach headline and every new regulation IMAGE, VIDEO, MUSIC AREN'T IN THESE NUMBERS. the $52M ceiling is text only. uncensored image and video are massive markets and venice has the only architecture serving them without retention. adult, creative, political, medical — entire categories where centralized AI structurally can't compete. completely excluded from the math above AGENT ECONOMY = $0 to minimal TODAY. x402 integration, no-KYC inference, native DIEM-based compute purchasing for autonomous agents. McKinsey projects $3-5T mediated by agents by 2030. venice has the only stack agents can actually use at scale — no KYC, no logging, no censorship. openai and anthropic can't follow without breaking the training data flywheel that funds them CRYPTO-NATIVE ECONOMICS THE PEERS DON'T HAVE. 69% of supply staked. burns now $2-10 VVV per new Pro sub (up from $1). emissions cut 8M → 6M → 5M, targeting 3M by july. demand sinks scale with users AND revenue (discretionary burns) SELF-FUNDED, ZERO VC. every peer above raised $300-500M . venice has no preferred stack to clear 6X VOLUME IN 4.5 MONTHS. faster growth than any peer at comparable ARR scale so the honest answer: after the rally, venice is fairly valued IF you compare it to openrouter, fireworks, and together AI. but those aren't the right comp set. the right comp is "privacy-moat inference platform with crypto-native economics, agent rails, multimodal expansion runway, and distribution still untouched." that company should trade at a multiple of those peers, not parity fair today. structurally undervalued forward caveats: - $52M API-equivalent is the only number directly observable from the data - the $30-50M range is reasoned from on-chain burns X commentary - chart is text only — image/video/music revenue would add to cash - DIEM is cash-flow negative for venice (deliberate design) - private comps are illiquid equity rounds, not liquid tokens - 75/25 input/output is an assumption tldr: the rally looks like a lot. the math says it's fair. the math also misses image/video, agents, distribution, and the privacy crypto-native combination nobody else has. fairly priced today on the wrong comp set. on the right one this is structurally still too cheap lmk if theres anything i missed / thoughts and questions

@TrustlessState appreciate the open question. Mine 👇🏻 Short version: $VVV isn't a token bet. It's an infrastructure bet on what private AI compute looks like at scale, with mechanics that compound rather than dilute. Long version, since you asked. @askvenice is the AWS of private inference. E2E encryption, no logs, 244 open-weight models, 2M users. They're billing ~$25K/day visible on-chain (just from new on-chain signups, which ignores renewals, API credits, and Enterprise). Real ARR is multiples higher: modeling SaaS retention puts steady-state at ~$170M . Of that ~$25K/day in observable revenue, ~$5K/day already converts to $VVV buybacks on-chain (programmatic Pro Sub TWAPs, depending on cycle phase). Lifetime: 29K buybacks executed, ~$606K USDC spent. The program runs, it's documented, and anyone can audit it. Now the part normies miss: $DIEM. $DIEM is a credit token for AI inference. 1 DIEM ≈ $1/day of queries on @askvenice, forever. A perpetuity of compute, on-chain. Target supply is ~38K, because every DIEM is an ongoing financial responsibility for Venice. They have to deliver that inference, indefinitely. But it works like a gym membership: not everyone uses what they pay for every day, and Venice procures inference well below retail. The functional cap on real costs is much lower than $38K/day. The kicker: supply already sits ABOVE the cap (~38.7K). People paying $1,629 per $DIEM to mint past the limit. That alone is a massive demand signal. How supply emerges, and why it gets harder: • To mint $DIEM you must lock $VVV on an exponential curve. Each DIEM costs more $VVV locked than the last. Mathematically intractable above a point. • To UNLOCK $VVV you must get $DIEM and burn it (some ppl have to buy it back lol). Sink in both directions. • Result: real free float of $VVV is ~29% of circulating ($752M mcap / $1.31B FDV). 69% staked, 27% of staked is locked. The flywheel 🔁 1. AI demand rises, $DIEM gains value (more inference, more valuable). 2. $DIEM more valuable, locking $VVV to mint becomes attractive. 3. Locking $VVV reduces $VVV float, $VVV up. 4. $VVV up, minting $DIEM gets even harder, $DIEM scarcer. 5. Unlocking $VVV many times requires buying burning $DIEM, buy pressure on $DIEM. Self-reinforcing in both directions. No marketing required. Pure mechanical design. Now the central thesis: AI agents. Picture an autonomous agent 2-3 years out. To be genuinely independent it needs: • Private inference (can't depend on OpenAI logging everything it does). • Predictable payment (doesn't want to swipe a credit card on every loop). • Perpetual access, no KYC. Solution: hold $DIEM. 1 DIEM = guaranteed daily inference budget, on-chain, private, forever. $DIEM is the first crypto-native primitive solving this cleanly. For an agent (or someone running agents heavily), $DIEM is literally the elixir of life. Not a figure of speech: it's what separates an agent that can operate continuously, indefinitely, autonomously, from one that needs a card swipe every other day. Apply that to the exponential curve every major lab is pricing in (Anthropic, OpenAI, Google, you name it) and $1.31B FDV in a world with billions of agents is probably very, very cheap. A bit like capitalizing Cisco in '95: not buying current revenue, you're buying the infrastructure of an exponential wave. Risks, since I'm not selling you only the upside: • Thesis rests on agent-economy growth at the scale markets are pricing. • $VVV is heavily staked locked, double-edged on sudden unlocks. • AI privacy regulation still uncertain (EU especially). • Liquidity still thin vs majors. Size accordingly. Bottom line: you're not buying a token. You're buying a thesis. If you don't believe in AI agents at scale, none of this clicks. If you do, the asymmetry is here: • Scarce token with asymptotic demand ($DIEM, cap already breached). • Mechanical sink token ($VVV, illiquid in both directions). • Real revenue, growing fast. • Thesis validated by every major lab. Up to each of you to draw a long-term price target from all this. Mine I'll keep to myself: saying it out loud would just get me called cooked. But that's crypto. 🤠🤙🏻 Every number above is verifiable in real-time on @venicestats. Yes, I built it. 🍷 (venicestats.com)

Ansem shares his thesis on $VVV “Venice AI is essentially a private inference platform… you can access all of the top state-of-the-art models, all the best open source models but through Venice, it doesn’t store any of your information” “They’ve done really well. They have over 2 million users. I think they have 50k daily actives on their app and they’re making money from the subscription” “One of the cool features of Venice that I thought was interesting is they have a second token that’s called DIEM… if you have 1 DIEM, you get to use $1 of compute per day for that 1 DIEM that you have” “It’s a really interesting model. I think it’s honestly attracted a lot of people to stake $VVV. We don’t have a lot of real protocols there in the crypto and AI space that aren’t vapor and are actually doing well”

As a large VVV staker I often don't use my full daily DIEM allowance. That's inference capacity sitting idle every day. So I'm building DIEMpool - stake your unused DIEM, let others pay to use the credits at a discount, earn yield. Waitlist: diempool.com @ErikVoorhees any chance this becomes native to @AskVenice? Would happily scrap this if you're planning it 👀

Can confirm. I’ve managed to do a LOT more for much less thanks to this GLM 5.1 update. And it’s private thanks to @AskVenice

Venice has quietly become the best aggregator for AI resources • Web app mobile API • Text/code/image/video • All frontier proprietary models • All leading open-source models • Pay with fiat or crypto One account, one API key, let's you access all of it. Built on principles of privacy, free speech, and user-sovereignty.

One thing nobody tells you about having kids is how much money you’ll spend on blueberries.

TEE and E2EE coming to a Italian city near you…

Venice LLM inference over time. Each color is a different model. Y-axis = billions of tokens per day.