AI is entering a new phase with the focus shifting to safety, ROI and sustainable economics. I joined @cnbckaren and @BenMBoulos on CNBC’s Squawk Box live from London after @Harnessio’s Unscripted conference. A few things we covered: 1) AI agents need boundaries. Agents are trained to accomplish a goal and will find different ways to get there. That means we need clear constraints around what they’re allowed to do. 2) More code doesn’t always mean more value. We’re seeing roughly 3x more code created with AI, but only a 25–30% increase in code shipped and driving business outcomes. Companies need to focus on the right model, context and task. The opportunity is to close that gap by focusing on the entire software delivery lifecycle, not just code generation. 3) AI economics have to work. Heavy model and inference costs are putting pressure on margins. Ultimately, AI companies will need sustainable unit economics. The next phase of AI will require more discipline around where and how we put these capabilities to work. Thanks Karen and Ben for having me! Full interview: piped.video/watch?v=HApdzxDS…

Today, LithosAI is launching a public preview of LithosBox: ultra fast versioned sandboxes that your agents can snapshot, fork, and restore in milliseconds. So far, we've been focused on providing ultra-low latency inference for open models like Kimi K3, DeepSeek V4.1, and GLM 5.3 Flash (stay tuned!). Now, we're giving agents runtime and post-training environments with Git-like control. With LithosBox, agents can: → spin up, branch, and snapshot thousands of environments as they explore multiple paths forward → decide on a winner and rebuild the winning state → then throw away the environments they don't need, all in milliseconds @dimitriosSkar, @JiaZhihao, and the rest of our team put together this agent vs. agent showdown with Doom so you can see how LithosAI helps your agents run 5x faster with AI serving infrastructure co-designed for ultra-low latency across inference and execution environments. In our demo, the agent completes its entire run 5.29x faster with LithosBox and Lithos DeepSeek V4.1 Flash inference than it does with leading sandbox and inference providers. Try LithosBox for free for up to two weeks via our public API.

Let's go @resolveai! 🚀

Introducing fal Recast, an easy way to change any character in one click. Upload an existing video + a reference photo of your new character. fal Recast changes who’s on screen while preserving the original motion, camera movements, cuts, and audio. Create character variations, adapt campaigns for different audiences, all without another shoot. Powered by H3 Max, the #1 model for overall quality, prompt understanding, and aesthetics against leading video models.

Portfolio News: @harnessio Announces Capabilities that Enable Security at Machine Speed. Harness introduced AI-powered scanning and agents that prioritize vulnerabilities, prepare fixes for review, and respond to zero-days. Virtual patching helps protect production while permanent fixes are deployed. Read more: harness.io/blog/harness-anno…

Portfolio News: @webflow opens agent access in @claudeai, @ChatGPT, and @cursor_ai, plus MCP access to Webflow Analytics. The company says agents can now work with a site’s design system and CMS, make changes on branches, and bring performance data into external workflows. Read more: webflow.com/blog/what-comes-…

Fal’s annualized revenue has doubled since March to $800 million, fueling talks for a new funding round at a valuation of $15 billion or more. Fireworks AI, which hit a $1 billion annualized revenue pace by July, could target a $30 billion valuation in its own potential round. More details: thein.fo/4xLemXN

LithosAI is pushing the Pareto frontier of agent inference speed, latency, and cost, and our latest results on Artificial Analysis proves it. Unlike most providers, offering more tiers improves our utilization. Read the full story:

September’s @Unusual_VC newsletter is out. This month: • Personal AI agents take off as Instinct and @Muse gain traction • @Anthropic releases Claude Fable 5.1 and Mythos 5.1, with upgrades across coding, agents and knowledge work • @OpenAI launches GPT-6 Astra, its most capable model yet • @Chalk introduces its MCP Server and Assistant for production ML • @Resolveai brings production context to coding agents • AI leaders call for tools to “pace the frontier” For more on enterprise AI and the unusual corners of it, subscribe: unusual.vc/news/

This summer, Penny Mares, our VP of Sales, had the chance to work alongside a particularly special Unusual intern: her daughter, Rose Foster. Turns out, the learning went both ways. A great reminder that the best mentorship is a two-way street. We loved having you on the team, Rose!

We're excited to release the State of Generative Media Report, Volume 2. Generative media has moved from experimentation into production and that shift looks different across every industry. Built from fal’s own platform data, the report explores how AI image, video, audio and 3D are evolving, written by the people at fal building the models, infrastructure and products behind the shift.

Designing a seasonal collection with 5 colorways per SKU? Region Maps lets you explore all of them in Vizcom without leaving the tool. Segmentation, editing, saving, and iterating, now all in Vizcom. 🎨🖤

AI in financial markets has a data problem. Last week, @Unusual_VC brought together AI leaders from @Nasdaq, @BankofAmerica, @jpmorgan and others to discuss “The Ontology Advantage” and why the missing layer in financial AI may be the ability to connect fragmented data without losing its meaning. A few themes stood out: • Industry classifications need to evolve with the economy • Connecting data isn’t enough, you need to preserve context • Domain expertise and high-quality data are critical for trustworthy AI The discussion was strong validation for what @TheiaInsights is building.

Security has always been core to Unusual Ventures. We recently brought together 50+ security leaders together to discuss what’s next. How do we govern AI agents at runtime? Do defensive workloads really need frontier models? And how can Blue Teams share intelligence at machine speed? The hardest questions often reveal the biggest opportunities.

Unusual portfolio company @harnessio and @kong are expanding their partnership to bring AI security directly into the infrastructure layer. The joint solution provides visibility and real-time protection across AI assets, agents, LLMs and MCP workflows. Read more: harness.io/press-and-news/ha…

FLUX 3 Video Edit is now live on fal - One prompt and the clip changes: swap a character, rebuild the background, restyle the shot, rewrite the dialogue - Lip sync follows the new dialogue automatically, including translations - Change colors, materials, and visual effects, or alter the events of a clip entirely

The valley is a very small place. Don't be transactional. Invest in building friendships with people you like. Be loyal and show up when things are difficult. I've known John for almost 20 years. It took 10 years from the first time he met for him to invest in one of my ideas but since then, he invested in 3 different rounds I raised and has been there anytime I needed some help, feedback or just to chat with someone.

LithosAI now has the highest speed, lowest latency, and lowest price inference for Kimi K3. And now you can test it yourself with our public API. Building coding agents, SRE workflows, or voice AI? Put LithosAI to the test: console.lithosai.cloud

What makes a founding team worth betting on? For Unusual’s John Vrionis: a technology inflection, deep customer empathy, domain expertise, rapid learning, and the conviction to pursue something fundamentally different. John has known @resolveai’s @spirosx for nearly 20 years and worked with him and @maynkag across three companies. We believe their technical depth, humility, and non-consensus conviction make them exceptional founders. We’re grateful to partner with them.

Unusual portfolio company @harnessio is expanding its partnership with AHEAD to help enterprises scale governed, agentic software delivery across the full SDLC. The expanded collaboration spans three areas: autonomous development and delivery, cloud migration, and security. Read more: ahead.com/news/ahead-and-har…