Instinct getting booted from the group chat (with my non-SF non-AI pilled friends) Multiplayer AI seems tricky… it’s not fun to come back to 50 unreads from an AI Assistant that someone else was prompting

We're bringing online 1 GW of net new compute in the next 24 months, and I'm hiring for Tactical Compute Ops in SF to help make it happen. You'll work directly with me and @nina_gerov, one of the best ops leaders I know. Finance/IB, Legal, and other backgrounds are more than welcome. DM me with a few lines on yourself and why you'd want to join.

The agentic web doesn't need a better search page. It needs search that's cheap enough to disappear into every task. On 20VC with @HarryStebbings, @paraga predicts agents will search 1000x more than people and says each query needs 10-100x less compute. Interesting idea that the real bottleneck might be retrieval economics, not model intelligence.

Our CEO and co-founder @burkaygur opened our GenMedia Conference in San Francisco this morning with one message: cost shouldn’t decide which ideas get made. We’re building every layer, from fal Agent and the H3 Max model family to the compute that runs them, so creatives, builders, and enterprises can make whatever they imagine.

Today, GenMedia Conference 2026 begins. 500 builders, creators and leaders shaping what comes next in generative media, all under one roof. Our CEO and co-founder @burkaygur kicks things off in a few minutes.

Data is the biggest bottleneck for agents now, not models. Parallel just made it a config line -- Crunchbase, Harmonic, Polymarket, Apollo, Similarweb, plus free PubMed and ChEMBL, live inside @p0's research APIs. Bring your own license or pay per call.

Despite what the billboards say, demand for great sellers (especially at hot AI companies in SF!) is at an all-time high. Products are becoming more technical and features are changing faster. Categories are becoming extremely crowded with new entrants pretty much every day, while buyers are harder to reach as they get overwhelmed with outbound slop. The easier it gets to build software, the more valuable it becomes to be great at selling it. For someone who’s early in their career or looking for a change, trying to get exceptionally good at that craft is a pretty interesting bet to make right now. We want to help more people make that move and break into sales even if they don’t have any previous experience, so today we’re launching @firstround Sales Combine. After two days of FREE training in SF, we’ll make direct intros to companies @NotionHQ, @clay, @fal, @togethercompute, @p0, @reductoai and many others. Apply by 10/3: And see a preview of the training sessions here:

The alpha for application layer AI companies has always been in figuring out the right shape of the product that makes tokens incredibly valuable in each domain. Really enjoyed this conversation with the @AnthropicAI team + @WisprFlow @pendoio founders on the future of building agents and how it's getting easier and faster to just focus on the product and take concepts from prototype to production!

Things Parallel (@p0) shipped this summer... • Search Turbo - web search for agents at 200ms median latency and $1 per 1,000 requests • Search Fast - near-frontier search quality at ~700ms average latency, also $1 per 1,000 requests • Entity Search - find people and companies in seconds using plain English • Monitor API GA - watches the web and pushes updates to your agent, with citations • Google Cloud integration - native web grounding inside Gemini Enterprise Agent Platform • Responses API - live-web research with citations, using the OpenAI SDK. Three-line switch • Free default web search for new OpenClaw installs • Index - lets publishers see how Parallel’s agents use their content and get paid for it • The full API suite on Google Cloud Marketplace, billed on your existing GCP invoice And Search Advanced ranked #1# when Artificial Analysis launched its Search Index in August. Hard to make this tweet short. @p0 team is relentless.

Most founders working on GPU infra have spent the last few years in the space. @cbondaatje’s obsession stretches back 10+ years. He was working on distributed compute before neoclouds were a thing, starting his own companies (including a 2014 idea for desktop GPUs while still at Harvard) and also joining others, like @CrusoeAI (as the 3rd eng hire and racking their first GPU rig by hand) and @hyperbolic_labs (as their founding platform engineer building out a 200+ cluster Kubernetes federation). As AI companies face down a serious GPU scarcity problem today, Christian’s rare depth of experience here sets him up perfectly to build @AranyaInc: the multicluster operating system that turns any bare metal into custom production-ready clusters in 48 hours. Aranya automates all of the moving parts in GPU orchestration (from individual kernel drivers to federation across clusters at unprecedented scale) so engineering teams can focus on building AI products instead of keeping clusters running. This team has quickly built something very impressive that’s used by customers with a very high bar. They now operate over $500M worth of compute and have already scaled up a top 3 inference provider. Proud to lead their $9M seed here @firstround.

🚀 Clearest sign of a rocketship Series A company is not enough bathrooms. The office felt big at first but the company grows too quickly. You can cram in more desks. But bathrooms are the bottleneck Saw this bathroom problem happen at fal and Parallel and now there's another company I work with experiencing this Could be alpha signal for growth investors We feel this pain at First Round too