The Physical World Fights Back
The Monday AI Signal: Feb 8–14
Hey,
Welcome to the first edition of Monday AI Signal. A new weekly series I am experimenting with, tracking pre-seed and seed AI deals globally to surface patterns before they become consensus.
February 8–14: Top Line Numbers
62 early stage AI deals | $306M raised | 20 countries | Median: $3.4M
The split: 46 seed and 17 pre-seed early-stage AI deals.
The number that matters: 20 of those startups are building for the physical world. Robots, haptic controllers, underground sensors, braided robotic bodies. Not another chatbot wrapper. Hardware.
For a market obsessed with foundation models, that’s a meaningful signal.
The Signal
Three early-stage AI trends stand out this week:
1/ Sixty percent of this week’s deals came from outside the US: It actually spanned 19 countries. San Francisco and London tied at 9 deals each. But the long tail is the story: Budapest, Dubai, Melbourne, Helsinki, Christchurch.
2/ The Seed infrastructure is scaling on all floors: The investor pattern this week tells a story on its own. Andreessen Horowitz showed up in 3 deals across 3 continents. Y Combinator appeared in 4.
But it’s not just the brand names...the seed infrastructure layer is moving in lockstep.
500 Global wrote checks in 2 deals totaling $20M
Antler backed 2
Seedcamp backed 2
Accel backed 2
Top-tier VCs and the pipeline builders are both deploying aggressively at the same time. We can assume the brand names will pick the winners…And the seed accelerators and funds build the funnel from which those winners come.
Both layers scaling up at once means the entire early-stage AI ecosystem is expanding beyond the headline rounds.
3/ More startups, smaller checks: Deal count jumped 38% week-over-week... but total funding barely moved (-0.4%). The median deal dropped 33% from roughly $5M to $3.4M.
Breakout Early-Stage AI Startups
The biggest pattern this week isn’t about software. It’s about startups betting that AI’s next chapter lives in the physical world.
1/ The Biological Computing Co. raised the week’s biggest round at $25M Seed, building inference hardware on biological substrates. At 7x this week’s median deal size, that check signals investor conviction that bio-compute is an infrastructure bet (not a research project).
2/ Haply Robotics raised $12M in Montréal for haptic controllers that let machines “feel.” Precision instruments for surgical training, warehouse automation, and drone piloting. Sound Media Ventures led, with Amazon’s Industrial Innovation Fund participating.
3/ Allonic pulled in $7M at Pre-Seed (backed by OpenAI and Hugging Face) for a process they call “3D Tissue Braiding.” Based in Budapest, it weaves robotic bodies from multiple materials in a single continuous process, the way biology builds organisms. Visionaries Club led.
4/ Stanhope AI in London raised $8M to bring neuroscience to autonomous systems. Co-founded by the researchers behind the Free Energy Principle, their AI adapts on the fly in unpredictable environments. This is the opposite of how LLMs work. Frontline Ventures and Paladin Capital led.
5/ Geolinks Services (based out of Paris) raised $7M to use ambient seismic noise for mapping fluid flow underground: applicable to mining, CO2 storage, and hydrogen exploration. Even the subsurface is getting the AI treatment now.
What connects all five: they’re building intelligence that operates under physical constraints, such as latency, power draw, and unpredictability.
The other half of this week’s standout deals share a different pattern: AI applied to problems so specific that only a domain expert would know they exist.
6/ Meridian, based in New York, raised $17M Seed (the week’s largest software round) for an AI-native IDE. They replace spreadsheets for institutional financial modeling. Not an Excel plugin. A standalone environment that integrates data sources, tracks logic flows, and automates model updates.
Andreessen Horowitz and The General Partnership led, with QED Investors, FPV Ventures, and Liquidity Ventures also in.
7/ Simple AI raised $14M Seed in San Francisco for voice agents trained on transcripts from top-performing human sales reps. The technology claims to handle thousands of simultaneous calls, understands context, and knows when to hand off to a human.
First Harmonic and Y Combinator led, with Samsung Next, True Ventures, and Conviction Capital also participating.
8/ Smart Bricks raised $5M Pre-Seed out of Dubai for an AI operating system that compresses 3–6 month real estate investment workflows into minutes: ingesting over a million data feeds to surface the top 0.1% of properties by risk-adjusted returns.
Andreessen Horowitz led, with Techstars, 500 Global, and Harvard Business School Alumni Angels.
9/ Seamflow raised $4M Seed in London for AI that automates the documentation, audits, and reviews that slow down certification timelines in the $300B testing, inspection, and certification industry. Starting in medical devices, now expanding to industrial infrastructure.
Northzone and Initialized Capital led. Mario Götze (yes, the footballer) is among the backers.
10/ Fluency raised $4M Seed in Melbourne for a work intelligence platform. Essentially, a browser plugin that watches how employees actually work, then auto-generates SOPs and identifies where automation would have the most impact. Accel led.
11/ ZeroDrift raised $2M Pre-Seed in New York for a real-time compliance firewall for regulated communications. It’s validating content against SEC and FINRA rules before it’s sent. Emerged from a16z Speedrun.
PS: How do you like this format? Whether it becomes a regular weekly series depends on you. So bother me with your feedback. What would you change? What more would you like to see? Hit reply!
Best,
Chintan


I really enjoyed this framing of “the physical world fights back” vis-à-vis AI. It really underscores that the next wave of AI value isn’t another chat interface but pushing intelligence into messy, constrained real‑world environments where latency, power, and uncertainty actually matter. The global spread of smaller, earlier checks plus all these deeply specific, non-obvious use cases makes this feel less like a hype spike and more like an actual build‑out of a new industrial layer.
The interesting tension now is this.
Are we at the beginning of an AI hardware cycle, or are these early bets before the economics of manufacturing bite back?