Architect (@Architect_Fi) header
Architect (@Architect_Fi) profile picture

The intelligence exchange. AI commodities futures exchange, dynamic inference markets, compute capacity and derivatives.

Chicago, IL
Joined September 2022
Photo by Architect (@Architect_Fi)Photo by Architect (@Architect_Fi)Photo by Architect (@Architect_Fi)Photo by Architect (@Architect_Fi)Photo by Architect (@Architect_Fi)Photo by Architect (@Architect_Fi)Photo by Architect (@Architect_Fi)Photo by Architect (@Architect_Fi)Photo by Architect (@Architect_Fi)Photo by Architect (@Architect_Fi)
Today we’re launching Liquid Inference, an OpenAI/Anthropic-compatible router where inference providers compete on price for every prompt. Buyers route across 700+ open- and closed-weight models at the lowest available marginal cost, providers onboard instantly to start serving models.
8
24
12
140
37,155
How the first OTC option on GPU rental prices works. @hyperbolic_labs bought a put on NVIDIA B200 rental prices. We designed it with them through our CFTC-registered Introducing Broker affiliate, and @luxor priced and executed it as counterparty. It settles in cash every month for a year against @silicondata’s B200 Rental Price Index (SDB200RT), on about $5M notional of rental revenue. Hyperbolic pays a premium for the option. In any month the index settles below the strike, it’s paid the shortfall on the covered volume. How compute hedges work: architect.co/otc?utm_source=…
3
8
1
31
3,073
Set rules on every request. One “exchange” field in any OpenAI or Anthropic call: a first-token deadline, no prompt retention, and no cut-off answers. Offers that can’t meet them don’t compete. architect.co/inference/docs/…
1
5
1
20
1,577
Proud to be the inference partner for @SuperteamTR’s university hackathon with @dev3pack. $20 of credits per team. Go build.
Big news for our university builders! ⚡ ️ Architect Inference by @architect_fi is the exclusive hackathon inference partner to our 4-week @dev3pack hackathon! 🤖 $20 in AI credits per team 🏆 Top 3 teams: $300 / $200 / $100 Join us 👇
1
4
13
1,203
With Anthropic’s IPO, earnings and disclosures will put public-company numbers on compute and inference markets for the first time. What we’ll finally see: • How tightly Anthropic’s cluster commitments track the earnings of the largest neoclouds and hyperscalers, i.e., how much of the GPU rental market is effectively just Anthropic • To what extent biggest labs are shifting from renting capacity to owning • The margins on the crack spread between raw compute cost and inference revenue, and how those compare to the same spreads for open-weight models • The sensitivity of lab revenue to macro and commodity correlates, such as electricity, memory, and metals prices • The trend of training time and cost for each successive frontier model pre-training Congratulations to Anthropic on a historic market event for super intelligence.
16
16
3
129
9,724
Most of an agent run doesn't need your best AI model, so we split it: Opus drives the main agent, subagents go to Auto and get filled from live offers on the exchange --cap 0.50 means no single request ever costs more than that
Run Claude Code on Architect Inference: arc login arc claude --cap 0.50 Opus runs the main agent. Subagents go to Auto, which buys from the offers on the exchange. No single request can cost more than your cap.
1
3
210
Run Claude Code on Architect Inference: arc login arc claude --cap 0.50 Opus runs the main agent. Subagents go to Auto, which buys from the offers on the exchange. No single request can cost more than your cap.
2
5
2
31
2,156
Congratulations to Architect’s partner @ComputeDesk on the round! Big things are coming from the Compute Desk team, and we’re excited to work with them on US compute futures, the first exchange-for-physical network, and building intelligence markets.
Announcing @ComputeDesk raised $45M, led by @lightspeedvp, to solve compute risk. I left Meta last year to build a team of quants and GPU infrastructure engineers to build the clearing infrastructure for compute.
9
11
1
90
4,398
Architect retweeted
Compute Desk raises $45M to build clearing infrastructure for compute
9
18
18
157
730,692
We just executed the first OTC option on GPU rental prices with @Architect_Fi and @luxor 🤝 A put hedging ~$5M of our B200 rental revenue against AI compute price volatility. Thank you @BrettHarrison, Matt Williams, and both teams!
6
8
4
37
7,680
Architect retweeted
excited & proud of @strobefund portco @Architect_Fi to print the first OTC options trade on compute -- helping an AI company hedge their future GPU costs reach out to @BrettHarrison & team if this or other compute-related trades are of interest
Announcement: Architect, @Luxor, and @hyperbolic_labs have executed the first OTC option on compute. Through our CFTC Introducing Broker we structured a bespoke $5M put option on B200 GPUs settling against @silicondata’s index. Luxor priced and was counterparty to Hyperbolic.
1
1
11
474
Today we're announcing the first OTC option trade on GPU rental prices. @hyperbolic_labs is hedging its NVIDIA B200 rental revenue with a put option designed by @Architect_Fi and executed by Luxor, settled against @silicondata's B200 Rental Price Index. luxor.tech/news/corporate-ne…
14
24
11
104
15,794
Architect retweeted
Compute is becoming a globally traded commodity. And every commodity market needs a way to manage price risk. Today, we executed the first OTC option on GPU rental prices with @Architect_Fi and @luxor, turning a year of B200 rental revenue into something we can plan around. A liquid compute market needs more than a way to buy and sell capacity. It needs the financial infrastructure to manage risk, too. Proud that Hyperbolic is helping build it. Thanks to @BrettHarrison, Matt Williams, and all three teams for making it happen. First of many.
We just executed the first OTC option on GPU rental prices with @Architect_Fi and @luxor 🤝 A put hedging ~$5M of our B200 rental revenue against AI compute price volatility. Thank you @BrettHarrison, Matt Williams, and both teams!
10
8
2
82
5,550
The first option on GPU rental prices just traded, and it settles on @silicondata's B200 Rental Index. Crude oil options started trading on NYMEX in November 1986. 40 years later, compute has its first one. Proud to be on the team building it.
Today we're announcing the first OTC option trade on GPU rental prices. @hyperbolic_labs is hedging its NVIDIA B200 rental revenue with a put option designed by @Architect_Fi and executed by Luxor, settled against @silicondata's B200 Rental Price Index. luxor.tech/news/corporate-ne…
2
2
22
1,843
Proud to be the first to hedge GPU rental revenue with an OTC option. AI compute prices are volatile, and now there’s a way to manage that risk. Thanks @BrettHarrison, @Architect_Fi, and @luxor for building this with us!
Announcement: Architect, @Luxor, and @hyperbolic_labs have executed the first OTC option on compute. Through our CFTC Introducing Broker we structured a bespoke $5M put option on B200 GPUs settling against @silicondata’s index. Luxor priced and was counterparty to Hyperbolic.
3
3
29
2,585
Announcement: Architect, @Luxor, and @hyperbolic_labs have executed the first OTC option on compute. Through our CFTC Introducing Broker we structured a bespoke $5M put option on B200 GPUs settling against @silicondata’s index. Luxor priced and was counterparty to Hyperbolic.
22
31
12
276
20,932
Architect is taking account of what our company has grown into this year. We have a new mandate and a new website. Dynamic inference markets, GPU capacity trades, OTC compute derivatives, US futures, developer tools, data aggregation, and more to come very soon.
16
18
4
126
9,190
We compared Liquid Inference with 8 routers and gateways. Each page has a head-to-head table, a fee calculator and dated sources. What the exchange does differently: providers compete for each request, and every request can have a hard cap. liquidinference.ai/compare
2
5
1
19
1,586
We've steadily improved our Liquid Inference routing algorithm to optimize for cache hit-rate. We've achieved >90% average across models, often realizing higher rates with DeepSeek and GLM providers.
16
11
68
4,442