2025 Year In Review

January 20, 2026

Hi Friends,

It’s 2026, and looking back at the last twelve months from the New York office, the vibe has shifted. 2025 was the year the industry stopped speculating about AI and started shipping it. We moved past the "can we build this?" phase and into "how do we scale this?”

Investor Dion Lim likes to use a wildfire metaphor for these cycles. It’s a bit intense, but it fits. The last year cleared out a lot of the companies built on hype, making room for the ones with real foundations. At Flybridge, we spent the year looking for those sustainable outliers.

Here’s a look back at 2025, and where we’re going next.

🗽 New York’s Ground Game

The idea that you need to be in San Francisco for deep tech is no longer true. 2025 proved that while the models are built on the West Coast, their real-world application is happening in New York.

Our advantage is density. In one afternoon here, a founder can grab coffee with a Columbia researcher, an NYU surgeon, and a Fortune 50 exec. That’s why Noetica is now being used by 17 of the top 25 law firms to manage over $1 trillion in deals.

A few things that kept me bullish on NYC last year:

  • PhDs: NYC and Boston are now graduating more AI/ML PhDs annually than the Bay Area.

  • Fortune 100s: 70% of the Fortune 100 are outside of SF, and many are in our backyard.

  • Empire AI: New York’s $750M AI investment is providing the kind of research anchor that Stanford once provided for Silicon Valley.

🚀 The 2025 Thesis: Agents and Efficiency

Our latest seed fund is focused on three areas where the tech is finally matching the promise:

  • Autonomous Agents: We’re evolving from tools that help people work to agents that do the work. Blitzy is a great example. They have 20 employees and 4,000 AI agents shipping millions of lines of code.

  • Real Infrastructure: As AI becomes the default, builders need better oversight. We backed Mako for GPU optimization and VoiceRun for human-like voice infrastructure.

  • Human Potential: BoldVoice is doing over $10M in revenue with less than 10 employees while helping tens of thousands of non-English-native speaking professionals build their confidence. Sunflower is providing addiction support to people who otherwise couldn't access it, crossing 10M sober days tracked now.

We’re also seeing a massive trend in efficiency. Our 2022 cohort is 2.6X more efficient at generating revenue than our 2019 cohort. This is continuing at an astonishing rate for our 2025 backed folks. Founders are reaching millions in ARR before they even look at a Series A.

🥳 2025 Highlights

  • My sobriety: I hit three years sober. Being fully present for my family is the most important thing I do, and it’s why I’m so proud to support Sunflower.

  • BrightHire Exit: Zoom acquired BrightHire. They survived the 2022 market crash, became lean, and doubled their revenue. It was a hell of a comeback.

  • OpenFX: They reached $32B in transaction volume in less than a year. It’s proof that stablecoins have a massive future in B2B foreign exchange.

  • Family: Charlie is a full-blown toddler now, and Holden remains my most honest advisor. He's also an expert in vibe coding with Replit.

😓 2025 Lowlights

  • The Burn: The wildfire metaphor has a downside. Some companies won't make it. Some will be acquired far below their peak. But that talent will get redistributed back into the ecosystem and will go on to create even more exciting things in the future.

  • The Divide: There is a clear gap now between businesses that are adapting and adopting AI and those that aren't. Watching friends end up on the wrong side of that line was the toughest part of 2025.

📌 5 Predictions for 2026

  1. The Jagged Frontier: AI is an "uneven genius." It can solve complex math but fail at basic logic. The winners will be the ones who solve for that inconsistency. AI can do so much, but just like humans, it can also make mistakes. How do we solve for that when it's an increasingly more important part of your work and your life every day?

  2. Vibe Coding: Appwrite is leading a shift where developers build apps through conversation rather than manual coding. Vibe coding is not a won battle yet. There's so much to be built here around security, defensibility, and distribution, but we know that software development will never be the same.

  3. Generative Engine Optimization (GEO): As AI agents start searching the web for us, companies like Limy are helping brands ensure they are the ones being cited. The future is unknown as to how ChatGPT and others will affect this, but we know it's a big problem to be solved.

  4. Momentum as a Moat: In an AI-first world, speed is everything. Moving faster than the competition is the only real way to protect a lead. It's not a long-term moat, but it's certainly one that you'll need to raise capital and to drive excitement in the early days.

  5. The Training Gap: The outlook on AI-driven unemployment is significant. The IMF estimates that nearly 40 percent of global jobs are exposed to AI. Similarly, the World Economic Forum predicts that while AI may create 170 million new roles by 2030, it will simultaneously displace 92 million existing jobs, making large-scale retraining an urgent necessity for the global workforce. We're watching this category closely and have already made one investment in a company that's currently in stealth.

📸 This Year in Photos

Thanks for being in my orbit. Let’s see what we can get done in 2026.

— Jesse