We first backed @cognition in 2024, when Scott and the team were building ahead of the frontier. @DevinAI's architecture anticipated capabilities the underlying models did not yet have. We believed those capabilities would catch up, and that the harder challenge would be turning raw intelligence into finished enterprise work. AI would move from local to cloud and individual to collaborative, with harness engineering coordinating models, agents and humans into reliable execution. That's increasingly what we’re seeing. Better models enable agents to operate over longer horizons and take on more complex, higher-value tasks, giving humans far greater leverage and broadening the range of enterprise work Cognition can reliably execute. We’ve continued to double down as the team translated a remarkably clear thesis into incredible commercial traction, while showing a level of talent magnetism that's hard to overstate. It feels like every day another engineer with their own lore shows up to build alongside @ScottWu46, @russelljkaplan, @stevenkplus1 and @walden_yan... Today it's @alexstamos. Welcome to the team! Our latest on Cognition and the future of machine intelligence: 8vc.com/resources/comparativ…

The next generation of American manufacturing won't be built by replicating yesterday's factories. It will require fundamentally rethinking how we translate hardware innovation into production at scale. Hardware products are evolving at an increasingly rapid pace, but the associated manufacturing processes remain a critical bottleneck. Every design iteration can require extensive engineering work to reconfigure assembly lines, slowing production and making domestic manufacturing prohibitively expensive. Led by @bobmcgrewai, a longtime friend of 8VC, @Arda_Industries is tackling this challenge by applying AI to manufacturing engineering. Already deployed at @anduriltech, the technology is dramatically accelerating the translation of product designs into production, laying the foundation for a future of increasingly autonomous, adaptable factories. We've long believed that some of the most consequential applications of AI will be in transforming the physical world. We're excited to back Bob and the Arda team as they redefine what's possible in advanced manufacturing!

Ramp just announced a $1.85B raise @ $60B; growing rev over 100% YoY, they're accelerating even faster at multi-B scale than before! For a masterclass in speed, obsessing over outcomes, killing bureaucracy, and crushing the competition, learn from @eglyman & the @tryramp team.

Today, we are announcing Nimble-v4 and Nimble-v4-max which are pushing the frontier of decision models, outperforming Jev and other variants. On the Decision Index public scores, nimble-v4-max and nimble-v4 score +10 and +6 against Jev 1.13, making nimble-v4-max the #1 model. Beyond benchmarks, we have done a lot of internal evals, and see meaningful improvements: 43% fewer mistakes than Jev, 50pp improvement on model routing decisions, being great at coding use cases, fact checking etc. The models are available via our API service. Go build nimbly: console.bespokelabs.ai

AI needs clear liability. Today, @harborai_ comes out of stealth with a $10M seed round led by @Base10Partners, @AbstractVC and @8vc. @elechi_v0, Cristian, Luis and I started @harborai_ with one mission: to insure the agentic economy. Rather than pacing the frontier, we believe the solution lies in pushing this frontier with real skin in the game.

Pit Stop: Pixel Eight x @tryramp Design. Williamsburg, Oct 22, 6 to 9pm. Part listening bar, part car showroom, part creative market. Come hang with the design community. Set a lap and RSVP 8vc.com/pixeleight

Military logistics shape battles, wars, and the course of history. We launched Gallatin from 8VC Build because our logisticians need precision on par with our weapons. Today, they’ve announced a $50M Series A to advance that mission. We’re proud to once again be a part of it.

We've announced our $50M Series A. Military logisticians deserve better than spreadsheets and endless briefings. Now we're putting Navigator in the hands of every allied sustainer. To the spreadsheets: we're coming for you. gallatin.ai/news/series-a-an…

NEW 🚨 Inside Anthropic — AGI, AI Optimism, Regulatory Capture, Open Source & More!  @JTLonsdale sits down with Anthropic's @_sholtodouglas and @the_marwell - Have we reached AGI — and what does it mean? - What are the most exciting breakthroughs at Anthropic right now? - What are the biggest risks and concerns? - What is their stance on open source? - Is Anthropic angling for regulatory capture? - Do AI researchers share same values as the heartland?  Episode 164 out now!  (00:00) Episode intro (02:00) Self-taught ML; Unlikely paths to Anthropic (04:30) What are the most exciting things you're working on? (08:04) Have we reached AGI — and what comes next?  (10:10) How to measure AI progress (14:20) Is biology the next great frontier? (16:50) Career advice & the decade of the generalists (19:30) What could go wrong? What are the biggest AI risks? (23:11) Offense vs defense; how technology evolves (25:08) Answering the regulatory capture critique (30:45) Anthropic's stance on open source (31:50) Would you slow down and let China go ahead? (33:05) Anthropic's view on distillation (40:12) Is a post-scarcity world achievable? (48:54) How do we preserve values, tradition & community? (55:00) Predictions for 2028

One of the hardest parts of enterprise automation is understanding how work actually gets done. @worktrace_ai maps real workflows to identify where automation can have the most impact. Now they’re closing the loop, going from discovery to executing the work itself. Already deployed across 20+ enterprises, with 75,000+ hours of automation opportunities surfaced. Proud to back this team!

AI has collapsed the timeline from vulnerability discovery to exploitation. @ArmadinSecurity is building for that reality with autonomous swarms of agents that continuously test customers’ own systems, simulating sophisticated adversaries to uncover exploitable paths and help security teams eliminate them before real attackers can. Just seven months out of stealth, Armadin is already in production with Fortune 500 companies and government customers. We’re proud to double down as the team announces its $255.5M Series B.

"What this enables is the return of American shipbuilding... We're going to create 10,000 jobs... and Port Alpha is going to be the most advanced, most efficient, highest throughput shipyard, not just in the country but really in the world." @Saronic CEO @MavrookasD on the significance of today's announcement. We launched Saronic out of 8VC Build in the fall of '22 with a small, incredibly talented team united around a bold idea: America could rebuild its shipbuilding capacity. Only four years later, they're building the ships and shipyards to restore American maritime strength. We are incredibly proud of this team. They're just getting started.

Earlier today, Vice President JD Vance helped break ground on Saronic's new Port Alpha, where a $3 billion investment is expected to generate up to 10,000 jobs and lead to an economic impact of $264.5 billion for the great state of Texas! American shipbuilding is BACK!

Looking good, gentlemen. Building and hiring in Texas to make America strong, thanks to the leadership of men like these!

Saronic has been awarded a production OT under the U.S. Navy's MUSV Marketplace for Marauder, our 180-foot dual-use autonomous ship — following a rigorous at-sea evaluation that tested the vessel’s capabilities. The trials included autonomously tracking and maneuvering around a variety of vessels, plus a battery of passive-perception tests — requiring Marauder to detect, track, and follow other vessels using only its passive onboard sensors, with no active radar. It's one of the hardest problems in autonomous maritime systems, and Marauder ran every scenario back-to-back on the same platform now under contract with the Navy. "Marauder went from initial design to on-the-water in under a year, and this evaluation demonstrated that the platform holds up under the Navy's own toughest testing," said Dino Mavrookas, CEO and Co-Founder. "We've now flipped our third Marauder hull, with several more moving through the production line behind it. We can build at that pace because we designed the autonomy, software, hardware, and the production process side-by-side, as one system, rather than bolting them together after the fact. That's what it means to be a vertically integrated shipbuilder, and it's why we've been able to build and deliver at speed and scale."

We first started investing in @cognition six financing rounds ago, in 2024; roughly eight years ago in AI years! What excited us then wasn’t simply the prospect of automating software engineering with some of the top talent we knew. @DevinAI pointed toward something much bigger, a true man-machine symbiosis where humans provide direction and judgment while agents execute and coordinate increasingly complex work across a jagged frontier of model capabilities. Two actual years later, the models improved remarkably quickly and the scope of what Cognition can do has expanded with them. We’ve continued to double down in every round as @ScottWu46, @russelljkaplan and the team have scaled with extraordinary urgency and ambition. Proud and honored to be along for the ride! Below is from @8vc on where we see machine intelligence heading:

Our case for Cognition (full piece with my colleagues @JTLonsdale @AlexKolicich @bela_becerra below) starts with two ideas from economics: Ricardo's comparative advantage and Coase's theory of the firm. Ricardo: gains from trade come from difference. Even a party that is better at everything comes out ahead by trading, because what matters is opportunity cost. A frontier model can be the best at every task and still be the wrong choice for most of the volume (every GPU-hour spent on routine edits is one not spent on work only it can do). The more jagged the frontier, the larger the gains, and the more it pays to be a good trading partner to every lab. A lab's own agent is protectionist by design (it buys domestic). Cognition trades with everyone and builds its own models where it has a comparative advantage. Coase: firms exist because markets are costly to use (finding a counterparty, negotiating, checking the work), so activity moves inside a firm wherever internal coordination is cheaper. As intelligence gets cheap, coordination becomes the scarce input. Ten thousand isolated copilots pay a transaction cost on every handoff, mostly in lost context. Devin brings those handoffs inside one system (shared context, planning, delegation, verification), with humans in the role Coase gave the entrepreneur: directing the work. Ricardo tells you where the surplus comes from, and Coase tells you what it costs to collect it. It goes to whomever can create the most intelligent system to solve this massive coordination problem and create true man-machine symbiosis (conclusion = build the best cloud agents). We heard a lot of skepticism when we invested at $2B and led the round at $4B. The labs were shipping their own agents, and the common view was that the layer in between would get squeezed. We thought better models would make it more valuable. Devin was a new kind of product, and it arrived early. It had an interface overhang (the mirror image of a capability overhang): you could delegate more to it than the models could yet deliver, and more than most teams were ready to hand off. The models grew into it, and cloud agents are now a category. The team built through that whole stretch like madmen. They are the most persistent group of people I've ever seen, and even by the standards of AI cycles, they move fast. Job not finished. 8vc.com/resources/comparativ…