Joined May 2019
A__Keith retweeted
Replying to @aphysicist
Really well put. Sadly, many don’t understand the difference and are actually raising money to run the factory. Thats a mistake. Reinventing the factory requires serious reinvention that has order of magnitude efficiency improvements (not just incremental improvements).
1
1
11
262
A__Keith retweeted
Replying to @tbpn
@GavinSBaker this is literally what @evanjconrad and @sfcompute are trying to solve for via a physically settled forwards market
1
4
305
A__Keith retweeted
SF Compute CEO @evanjconrad reveals why some AI startups are being asked to pay up to 5 years of GPU compute costs upfront just to secure computing power: "Many new clusters are being asked for investment-grade credit, which startups don’t have. So to make up for that fact, you’re instead asked for a big prepayment." "Now it’s not the first month up front, it’s 100% prepayment. Somewhere between 30 and 100%. We’ve seen 100%." "You are asked to pay for a three to five-year contract on a supercomputer with all the cash up front. That’s a death knell for startups." @sfcompute
Replying to @chamath
Pretty much. As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers. SpaceX closed $40B of debt for GPUs, but we’re investment grade and have Starlink & launch revenue. That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many.
3
2
19
4,500
A__Keith retweeted
Catch me on MTS this afternoon!
2
1
17
938
👇
Our ceo and founder @samanfarid on the best way to get to general purpose robots that don't suck. Thanks @AnneliesGamble for the collaboration! Read it:
1
122
A__Keith retweeted
It was fun to document what we've been learning about building robot deployments that ACTUALLY work, and what the path to true improvement is with @haomiaoh -- backed by experience deploying robots in hundreds of factories, we now know the balance to strike between generalizability and performance.
Our ceo and founder @samanfarid on the best way to get to general purpose robots that don't suck. Thanks @AnneliesGamble for the collaboration! Read it:
2
2
21
1,182
A__Keith retweeted
Super excited to share that the first episode in our new @weaviate_io webinar series is live on YouTube! 🍿 I loved learning more about the Weaviate MCP and @alexledbetter15's Delegance Brokerage! 🔥 It was great to share an update on some recent updates to Weaviate's Query Agent, including a new MCP server! 🚀 I hope you find it interesting! 💚👇 youtube.com/watch?v=VZjicnVS…
1
7
1
14
1,266
A__Keith retweeted
neocloud sells gpus to inference providers who sell to model routers who sell to ??? where is all this capacity for OSS models going to go without products that create demand someone's gotta make the apps
86
21
9
786
55,245
this event has expanded beyond what we can fit, if you have a venue we could borrow on thursday, we would be deeply appreciative
we have gotten a flood of questions from startups & venture firms asking how to protect themselves from the debt markets freezing. to coordinate, we're hosting a small off-the-record q&a in sf on thursday night you can sign up here: forms.gle/hg9RsJh2gKdNgkfW7
5
3
54
5,845
A__Keith retweeted
Exactly the things Formic has been doing for the last 5 years
there is no ChatGPT moment for robotics capability can skyrocket “overnight” but adoption is the long horizon all the opportunity belongs to those who can make adoption work. its a game of march of 9s (reliability), the last mile (deployment), and keeping it running (support).
1
2
25
888
we have gotten a flood of questions from startups & venture firms asking how to protect themselves from the debt markets freezing. to coordinate, we're hosting a small off-the-record q&a in sf on thursday night you can sign up here: forms.gle/hg9RsJh2gKdNgkfW7
2
19
3
151
21,604
A__Keith retweeted
SFCompute is not a neocloud SFCompute is not a neocloud SFCompute is not a neocloud It’s a compute market that also operates and handles physical settlement of clusters. This is VERY HARD to do.
SF Compute exists to make compute accessible for start ups while supporting larger customers. We do this through our underlying orderbook, which allows larger tenants to resell their idle capacity for the little guys to sublease and use. We invented compute markets through this process. Our orderbook makes us a forwards market that’s physically settled, where physical settlement means we design and operate the clusters behind the orderbook. You can actually trade physically-settled block compute on our platform today. We’ve been in the space since 2023. A lot of people have asked what we’ve been doing for the last 3 years and why it had taken us so long to emerge as a compute market. It’s actually extremely hard to do physical settlement. The plethora of compute markets you’ve seen popping up on your feed lately are not physically settled. SF Compute is not a neocloud. We could have made so much money by being extremely early in the space but decided not to, in order to create a new type of business in the compute industry so that start ups can continue to exist. You can get GPUs on both long term and short term contracts, you can resell what you don’t need to lower your overall compute price, you can block trade compute on our platform, and if you wanted to own the hardware yourself for its residual value, we can operate it for you as a service. SF Compute is everything you could want in one company. We’re kind, earnest people who reinvent the wheel. We’ll never try to sell you something we’re not proud of. We always try to give the very best for all the people we work with.
6
6
121
13,053
If you’re a venture firm looking to own your own cluster, san francisco compute offers supercomputing as a service. We’re an independent operator that lets you own the residual value of the cluster, so you can make this trade directly.
Heard from a neolab friend: “Even if we fail to train the best model, we locked in GPU capacity early. H100/B200/B300 rental prices have since gone up 2–3x. We can just resell the compute and print money.” OK, so that’s the business model.
3
3
4
86
7,193
A__Keith retweeted
Just end my day at AGNTCon/MCPCon here in San Jose. Spoke to so many people and had an amazing time taking with a ton of developers building agents and needing durable execution as essential for long running solutions. And most said "No SaaS, I want BYOC or full disconnected" I also walked the floor and counted the sponsor lineup. ~16 of ~110 sponsors are MCP/AI gateways: Solo, Kong, Gravitee, Traefik, WSO2, TrueFoundry, Runlayer, Obot, MintMCP, Stacklok, Airia, Cequence. That’s the most crowded single category. That one will consolidate fast. They are generally useful, but in the long run are a temporary security solution, in the light of agent identity being the required solution. The rest of the map: 🔐 Agent identity & security, ~14: Auth0, Keycard, WorkOS, Descope, Aembit, Chainguard 🔌 MCP connectors & tooling, ~12: Workato, CData, Boomi, Merge, Apify, Speakeasy 📊 Observability, ~9: Datadog, Elastic, Chronosphere, groundcover 💸 FinOps & monetization, ~7: Finout, PointFive, Stigg, Revenium 🤖 Agent frameworks & durable execution, ~12: AWS, Strands, Google ADK, Mastra, Temporal, Orkes, Diagrid The gateway layer is crowded. Which other categories consolidate?
4
2
6
341
If you do become a customer of @unkeydev I might just show up in your repo and make things better
3
3
2
25
2,223
Noting here that this is no shade on Modal; SFC has explicitly been using Modal as the prime example of a customer we exist to support (even if we're not there yet). We love Modal & we're explicitly trying to make it possible for Modal to sell you those 1k B300s.
1
1
2
223
A__Keith retweeted
NEW!! GOP Sen. Kevin Cramer says he received a briefing today from Air Force Sec. Troy Meink on Starshield, raising concerns over U.S. reliance on SpaceX. “Fighting wars” can’t be so dependent on private companies, he says.
82
611
65
2,927
390,812