Five AI Startups Betting on the Next Wave

The post-ChatGPT funding gold rush is not slowing down. If anything, capital is concentrating into fewer bets with larger checks. Over the past two weeks, five startups in particular have cut through the noise with substantive funding rounds, real traction, and technical approaches that move beyond simple chat wrappers.

Here is what they are building and why each matters.

Hark: The $6 Billion Bet on a Universal Interface

Brett Adcock does not build small. The serial entrepreneur behind Figure.AI and Archer Aviation launched Hark in late 2025 with $100 million of his own money. On May 21, the company announced a $700 million Series A led by Parkway Venture Capital that values Hark at $6 billion post-money [1].

The specifics of what Hark is building remain tightly guarded. The company describes it as a “universal” AI interface (models plus hardware) for personal assistance. First multimodal models are expected this summer, with dedicated hardware to follow. Investors include Nvidia, AMD Ventures, Intel Capital, Salesforce Ventures, and Qualcomm Ventures. Abidur Chowdhury, a former Apple product executive, leads design.

Hark currently operates a data center with Nvidia B200 GPUs and employs 70 people. The scale of the round signals investor conviction that the first true AI-native consumer platform will not emerge from OpenAI or Google but from a team willing to own the entire stack.

IrisGo: Andrew Ng Backs a Desktop Agent That Learns by Watching

IrisGo closed a $2.8 million seed round earlier this year led by Andrew Ng’s AI Fund, with additional backing from Nvidia and Google. The company is building a desktop companion that observes user workflows, learns them after a single demonstration, then automates them proactively [2].

Co-founder Jeffrey Lai is a former Apple engineer who helped build the Chinese language version of Siri. (The company name, IrisGo, is a not-so-subtle nod: “Iris” is “Siri” reversed.)

The software runs a hybrid architecture. Routine tasks process on-device for privacy; heavier workloads move to the cloud only with explicit user authorization and end-to-end encryption. A built-in skills library covers email drafting, invoice processing, report generation, and document summarization, while a coding assistant targets developers.

IrisGo recently launched macOS and Windows beta apps and struck a preinstall deal with Acer. The company is actively pursuing similar partnerships with additional laptop manufacturers.

NanoCo: Open Source Agents With a Security-First Stack

NanoCo, the commercial entity behind open source AI agent harness NanoClaw, raised a $12 million seed round led by Valley Capital Partners. Strategic investors include Docker, Vercel, Monday.com, and Hugging Face CEO Clem Delangue [3].

The company was founded by brothers Gavriel and Lazer Cohen. Gavriel wrote the first lines of code on his couch and, within six weeks, had a term sheet and a roughly $20 million acquisition offer. The brothers declined the buyout.

NanoClaw’s core differentiator is architectural. Unlike its predecessor OpenClaw, which runs with broad system access, NanoClaw agents execute inside Docker MicroVM sandboxes. API credentials never reach the agent directly; outbound requests pass through a Rust gateway that enforces policy and requires human approval for sensitive writes. A crowd of 300 people once chatted simultaneously with Gavriel’s personal agent during a live demo in Singapore. The agent safely rejected malicious requests while allowing legitimate calendar bookings.

The open source project has crossed 250,000 downloads and nearly 29,000 GitHub stars. Singapore’s Foreign Minister, Vivian Balakrishnan, publicly called it his “second brain.” NanoCo monetizes through managed enterprise deployments and forward-deployed engineering support.

Clouted: AI as a Penetration Tester for Social Media Algorithms

Clouted emerged from a16z’s Speedrun accelerator in 2024 and recently closed a $7 million seed round led by Slow Ventures. The platform automatically produces short video clips from longer content, then uses AI to test distribution strategies across social platforms [4].

The approach borrows from cybersecurity. Rather than treating social algorithms as black boxes, Clouted’s system runs thousands of experiments (different cuts, captions, posting times, and platforms) to identify what triggers viral distribution. Each campaign feeds data back into the model, making the next one more precise.

The company taps a network of over 100,000 gig creators for editing work, then layers AI on top for strategy. Founder Justin Banusing first stress-tested the technology promoting &Friends, a Manila-based electronic music festival that now draws over 20,000 attendees.

Resolve AI: When AI-Generated Code Breaks Production

Resolve AI sits at the intersection of two colliding trends: the explosion of AI-written code and the operational crisis that follows when systems break in production. The company raised a $125 million Series A at a $1 billion valuation earlier this year, led by Lightspeed Venture Partners with Greylock participation [5].

The platform deploys teams of specialized agents (not single bots) to investigate production incidents. Each agent pursues a different hypothesis, verifies evidence independently, and constructs full causal chains from root cause to symptom. The company reports a twofold improvement in root cause accuracy over earlier single-agent architectures.

CEO Spiros Xanthos previously co-created OpenTelemetry, the open standard for observability data collection. Resolve AI sells on an outcome-based credit model: customers pay only when an agent successfully troubleshoots an alert. Named customers include Coinbase, DoorDash, Salesforce, MongoDB, and Zscaler.

What Unites Them

None of these companies are building chatbots. Hark is building an interface layer. IrisGo is automating desktop workflows. NanoCo is sandboxing agent execution. Clouted is reverse-engineering algorithmic distribution. Resolve AI is debugging the mess that AI coding tools create.

The common thread is depth. Each startup addresses a specific, high-friction problem rather than wrapping a GPT-4 API call in a prettier frontend. That specificity is exactly what separates funding rounds that close from those that do not.

Sources

[1] Fernholz, Tim. “Hark raises $700M Series A for its secretive ‘universal’ AI interface.” TechCrunch, May 21, 2026. https://techcrunch.com/2026/05/21/hark-raises-700m-series-a-for-its-secretive-universal-ai-interface/

[2] Ropek, Lucas. “IrisGo, a startup backed by Andrew Ng, looks to become the AI desktop buddy you never knew you needed.” TechCrunch, May 20, 2026. https://techcrunch.com/2026/05/20/irisgo-a-startup-backed-by-andrew-ng-looks-to-become-the-ai-desktop-buddy-you-never-knew-you-needed/

[3] Bort, Julie. “NanoClaw creator turns down $20M buyout offer, raises $12M seed instead.” TechCrunch, May 20, 2026. https://techcrunch.com/2026/05/20/nanoclaw-creator-turns-down-20m-buyout-offer-raises-12m-seed-instead/

[4] Temkin, Marina. “Clouted wants to take the guesswork out of making short videos go viral.” TechCrunch, May 20, 2026. https://techcrunch.com/2026/05/20/clouted-wants-to-take-the-guesswork-out-of-making-short-videos-go-viral/

[5] Nuñez, Michael. “Resolve AI says the AI coding boom is breaking production systems. It wants to fix that.” VentureBeat, May 21, 2026. https://venturebeat.com/ai/resolve-ai-says-the-ai-coding-boom-is-breaking-production-systems-it-wants-to-fix-that/