NFX

Starcloud is excited to announce that we will integrate 50 SpaceX Starlink Mini Laser terminals across 25+ satellites.  Each of the Starcloud satellites will carry two Starlink Mini Laser terminals, and the first hardware is expected on orbit within one year. Starlink Mini Laser terminals, the same laser crosslink technology that SpaceX developed for its Starlink constellation, provide up to 25 Gbps of continuous intersatellite connectivity at distances up to 4,000 km and are capable of higher link speeds at shorter distances. The terminals enable direct optical links between Starcloud satellites and the Starlink constellation using laser light, eliminating the need for Starcloud to send data directly through bandwidth-constrained ground stations.

AI will create much more than it "destroys." It sounds like hype. Until you understand the pattern. History shows it. Art shows it. How technology expands markets with Pete Flint:

🚀 Another huge, set to be category winner, that started as a first check we wrote as part of our campus FAST program Blitzy Anna Piñol While there are an increasing number of programs that back student founders on campus, we remain highly selective (just like the rest of our portfolio) on who we work with and support - and its a huge testament to the Blitzy team for how far they have come in just 2 years. And speaking of our campus initiatives -- we're recruiting for our next batch of student Fellows, who work directly with us on the investment team, and invest directly in student companies like Blitzy through FAST. Apply here if you're currently a student at Harvard, Stanford, MIT or Berkeley (or incoming), Undergrads and Grads student welcome! Apply by May 15th! https://lnkd.in/ebqJ7eXe

OpenAI just made intelligent, real-time audio models available in their API. The infrastructure is ready. The question now: what do you build? We see 4 on-ramps for voice AI – ways for startups to translate this progress into product: ◼ Expansion of "AI as Labor" playbook — replaces labor in B2B at scale ◼ Conversational Trojan Horse — high-ROI voice solutions unlock platform expansion ◼ Conversational Data Goldmine — trusted conversations generate qualitative data at scale ◼ The Next Frontier — deeply human experiences that unlock entirely new markets Anna Piñol and Valerie Osband Mahoney break down each in detail: https://lnkd.in/eyeHbiZU

AI can only solve biology as fast as we can measure it. It's a massive constraint. Pumpkinseed just raised $20M to remove it. Their approach: reading protein sequence and structure directly from molecular vibrations. It's a new form of biological measurement that unlocks data previously impossible to capture. This closes the gap between real-world biological samples and the digital input AI models need to improve health at scale. We are thrilled to back this team during a critical moment for bio. Congratulations! 

Today we're excited to announce Blitzy's $200M growth round at a $1.4B valuation. To the teams at Northzone, PSG and Battery Ventures…thank you for your conviction to join us on this journey. Almost three years ago, while the industry raced to train larger models and supercharge developers inside an IDE, Brian Elliott and I saw a different path: fully automating enterprise software development. At NVIDIA, I had a front-row seat to AI applied to engineering at scale being in Jensen’s orbit. The lesson was that raw model capability does not solve hard engineering problems. You need a system around the models to drive predictable, reliable, secure outcomes. The real problem in the enterprise is not generating code. It is producing software that can actually be pushed to production. That means understanding and respecting the systems you are working on, orchestrating thousands of agents across interconnected systems, validating every output autonomously, and running across long time horizons without losing coherence. Two things set our approach apart: Reverse engineering before generation - Every project is validated against a knowledge graph of the existing codebase: architecture, dependencies, conventions, constraints. This grounding is what makes autonomy safe at enterprise scale. Orchestration at hyperscale - Most of the industry treats the model as the product and orchestration as plumbing. We inverted that. Thousands of agents work in parallel, bound by a pre-approved spec. Orchestration is where coherence at scale and in parallel is unlocked. To our customers, team, partners, and investors: thank you for backing the long road when these bets were not yet consensus. Finally, thank you to Forbes and Sofia Chierchio for telling our story: https://lnkd.in/eEMteG7s

A startup just built an AI system that outperforms OpenAI, Anthropic, and Google at real-world enterprise software development Now Blitzy is announcing a $200M raise at a $1.4B valuation. We were the first money in. Here's why we backed them at day 0↓ It comes down to 3 things: ◼ Contrarian thinking on architecture ◼ Reframing the market (AI services are the new software) ◼ The right founder pairing for the customer Here's their story from our POV: https://lnkd.in/e7Vq7REz

Anthropic ran an experiment last week that tells us something important about how marketplaces will evolve... It's called Project Deal. Here's what they found and what it means for founders: The setup: 69 employees. $100 each. One Claude agent per person. The agents interviewed each participant, learned their preferences, and were deployed into a live Slack marketplace. Result: 186 completed deals. $4,000+ in real goods exchanged. Agents posted listings, fielded counteroffers, and closed transactions entirely in natural language. Three things this tells us: 1. The search box is being replaced. AI is shifting commerce from "pull" (keyword search) to "push" — agents that know what you want and act before you ask. 2. New supply will emerge. One agent negotiated a day with someone's dog. That sounds like a quirk — but when friction disappears, new categories of things become transactable. We've seen this before. 3. Every platform shift opens a window. GPS + mobile gave us Uber. SEO gave us Zillow. AI agent marketplaces are the next unlock — and 46% of participants said they'd already pay for this. We wrote about what this means for founders building in this space — what the AI-first marketplace looks like, and where the durable value will accrue. Tactics for Building AI-First Marketplaces from Pete Flint: https://lnkd.in/efxrAaH5