# Humble’s $24M Freight Bet, PEARL’s Protein Model, and New AI Traction from Eurasia

*By VC Tech Radar • July 2, 2026*

Humble Robotics' $24M raise, Genesis Molecular AI's PEARL results, and a cluster of revenue-bearing AI startups from 500 Global's Eurasia batch stand out this cycle. Also notable: renewed robotics sentiment, a nuclear-powered AI data-center concept, and investor appetite for new Series A funds.

## 1) Funding & Deals

- **Humble Robotics raised about $24M led by Eclipse.** The company emerged from stealth in April and is building a fully autonomous cabless electric hauler for short-haul freight, with initial routes under 50 miles. The product thesis is deliberately constrained: CEO Eyal Cohen describes Humble as a search for the simplest possible robotics platform to move freight from A to B. Cohen says he has spent 20 years in Bay Area deep tech across 7-8 startups, and Humble reassembled veterans from seven previous companies. [^1]

- **Valar Atomics and Nvidia are partnering to explore a 30 MW nuclear-powered AI data center in Emery County, Utah.** The concept pairs microreactor energy with next-generation cooling designed for near-zero water consumption. The announcement is notable because it specifies both the power source and cooling architecture upfront. [^2]

## 2) Emerging Teams

- **Cerberus.** Aziz says Cerberus built an autonomous AI ethical hacker on a new programming language where every instruction must receive a mathematical proof of safety. He says he made a programming-language safety breakthrough at 17, runs Central Asia's top cybersecurity firm, has secured 300+ projects in three years, and Cerberus has already generated $150k+ in revenue with 15+ bank pilots. [^3]

- **Klinova.** Maryam Haraz Ashvili is targeting a real clinical-ops bottleneck: matching patients to trials across handwritten, multilingual, and legacy records. Klinova says it found 4,000 eligible patients for a recent migraine trial in two hours, launched six pilots across 1 million patient records, and signed 22 LOIs worth $1.2M in five months; the team includes an operator who ran 35 clinical trials across Eastern Europe and a second-time founder. [^3]

- **mybots.** The company says its AI sales agent can replace the CRM and outreach stack by finding customers, qualifying leads, sending offers, and closing deals. It reports $240k ARR after seven months, 25% MoM growth, and a 23% close rate across 147 businesses; the team includes a second-time founder and a competitive-programming champion CTO. [^3]

- **Tezbur.** Umid Ahmedov says Tezbur's AI-optimized hybrid delivery network has processed 600k+ parcels, reached $1.7M ARR, and is growing 40% MoM. The company says it has 500+ customers, partnerships in Kazakhstan, paying clients in Denmark, and recently acquired a delivery company in Uzbekistan that should double MRR from next month; Ahmedov previously led data and AI architecture at Microsoft. [^3]

- **Bloomy.** Y Combinator highlighted Bloomy as a mastery-learning platform for K-12 English, Math, and Writing. The product uses a Socratic AI tutor for standards-aligned instruction and teacher-facing insight tools, and early pilots showed about 2x faster than projected growth on NWEA MAP assessments. [^4]

## 3) AI & Tech Breakthroughs

- **Genesis Molecular AI / PEARL.** Co-founder Evan Feinberg and CTO Sergey Edunov are building a diffusion-based protein-ligand co-folding model that adjusts both ligand placement and protein structure, targeting induced fit and protein flexibility. In a recent OpenBind evaluation on 802 never-before-seen EV-A71 co-complexes, Edunov said PEARL was especially strong at moving the relevant loop; Feinberg argues that roughly 1Å RMSD is the threshold that really preserves interactions. Genesis says these gains have pushed its internal SAPPHIRE system toward agentic drug-discovery loops that can inspect poses, form hypotheses, read literature, and generate the next candidates. Edunov previously led Llama 2 training and Llama 3 pretraining at Meta. [^5]

> "If your model is sitting at 1.8, 1.9 Angstrom RMSD, that's slop, most likely." [^5]

- **Humble's autonomy stack is a concrete example of recent vision-model gains changing robotics design.** Cohen says pre-trained vision models quickly handled scenarios that used to require months of engineering work, including cones, traffic lights, and an officer holding a stop sign. Humble still uses lidar, radar, and camera on the vehicle, but says recent improvement in camera-centered vision models has been strong enough to reshape the stack. [^1]

- **Devin Security Swarm points to a more structured agent architecture for software security.** Cognition introduced it as a more cost-effective and accurate way to find vulnerabilities in complex codebases, based on an architecture it calls Agentic MapReduce. Harrison Chase's follow-up is useful because he ties the idea to a broader pattern: programmatically spawning subagents to get more deterministic control over how agents are created and run. [^6][^7]

## 4) Market Signals

- **500 Global's Eurasia batch is producing revenue-bearing AI companies across multiple verticals.** In one demo day, Cerberus reported $150k+ in revenue, mybots $240k ARR, May Call $28k MRR, and Tezbur $1.7M ARR across cybersecurity, sales, debt collection, and logistics. [^3]

- **Autonomous vehicle and robotics sentiment is warming again, but around teams that stayed through the last cycle.** In the TechCrunch discussion, the sector was described as starting to have a moment again in 2026, and Humble's CEO said many of the same operators from 10 years ago kept at it and are now beginning to see the results. [^1]

- **AI data-center design is increasingly bundling energy source and cooling architecture from day one.** The Valar Atomics/Nvidia concept pairs microreactor power with next-generation cooling designed for near-zero water consumption. [^2]

- **Some investors are explicitly framing new Series A funds as a source of alpha.** One current example is Ashton Kutcher leaving Sound Ventures to start a new firm with former NFX GP Morgan Beller; Garry Tan then described new Series A funds as probably the biggest alpha in all of VC. [^8][^9]

> "new Series A funds are probably the biggest alpha in all of VC" [^9]

## 5) Worth Your Time

- **[Latent.Space: The Coolest Diffusion Research Isn't in LLMs](https://www.latent.space/p/the-coolest-diffusion-research-isnt)** — the strongest long-form technical item in this set for PEARL, induced fit, and why diffusion primitives matter in 3D structure prediction. [^5]

- **[TechCrunch Equity on Humble Robotics](https://www.youtube.com/watch?v=OPiqNwy_45s)** — the clearest source here on the narrow freight thesis and why recent pre-trained vision models changed the autonomy stack. [^1]


[![Autonomous vehicle hype is back, and Humble Robotics is bringing it to freights | Equity Podcast](https://img.youtube.com/vi/OPiqNwy_45s/hqdefault.jpg)](https://youtube.com/watch?v=OPiqNwy_45s&t=1070)
*Autonomous vehicle hype is back, and Humble Robotics is bringing it to freights | Equity Podcast (17:50)*


- **[500 Global in Eurasia Demo Day](https://www.youtube.com/watch?v=KVhX7i_mQTg)** — a dense sourcing list for early traction across AI security, clinical operations, sales automation, debt collection, and logistics. [^3]

- **[Cognition's Devin Security Swarm thread](https://x.com/cognition/status/2072368168182432109)** and **[Harrison Chase's follow-up](https://x.com/hwchase17/status/2072377816780624266)** — worth reading if you are tracking agent architectures built around programmatic subagent creation. [^6][^7]

- **[YC on Bloomy](https://x.com/ycombinator/status/2072375624971551108)** — short, but one of the cleaner pilot outcome metrics in this set for AI tutoring. [^4]

---

### Sources

[^1]: [Autonomous vehicle hype is back, and Humble Robotics is bringing it to freights | Equity Podcast](https://www.youtube.com/watch?v=OPiqNwy_45s)
[^2]: [𝕏 post by @AtomsNotBits](https://x.com/AtomsNotBits/status/2072424019480170793)
[^3]: [500 Global in Eurasia, Batch 10 Demo Day](https://www.youtube.com/watch?v=KVhX7i_mQTg)
[^4]: [𝕏 post by @ycombinator](https://x.com/ycombinator/status/2072375624971551108)
[^5]: [🔬 The Coolest Diffusion Research Isn't in LLMs — Evan Feinberg & Sergey Edunov, Genesis Molecular AI](https://www.latent.space/p/the-coolest-diffusion-research-isnt)
[^6]: [𝕏 post by @cognition](https://x.com/cognition/status/2072368168182432109)
[^7]: [𝕏 post by @hwchase17](https://x.com/hwchase17/status/2072377816780624266)
[^8]: [𝕏 post by @ychernova](https://x.com/ychernova/status/2072121547746144534)
[^9]: [𝕏 post by @garrytan](https://x.com/garrytan/status/2072289609804128549)