focal.vc
Southbound W27 applications are open3 months in Miami for technical foundersApply

❯Your first lead.

We exist to put moonshot founders in business.

A firm built to back technical founders at inception.

Your first lead

We back technical founders at inception with speed and conviction. The more ambitious the idea the better.

50+
first rounds led to date

We designed our firm to be your catalyzing first partner.

75%+
within a month of incorporation

When we say nothing's too early, we mean it.

~10
investments a year

Every check is a core position. No option buying, no discovery checks.

Partner to moonshot founders

We back people first and foremost. They all have three things in common:

  • Obsessedwith pursuing the most asymmetric opportunities (no small ideas).
  • Technicalbuilders who ship with speed and intensity (at least one on the team).
  • Hungryto win, be challenged, and be better tomorrow than today.

That leads us across industries and verticals with an open mind. Only two non-negotiables:

  • It's your first financing
  • Your startup is based in the US or Canada

When we decide to partner, we move fast, often going from first meeting to investment in a few days.

open portfolio.md →

Your VC of record

focal is Pascal, Sydney, Daniel, and Amine.

Pascal, Sydney, Daniel, and Amine of focal

We are founders of our own firm and run focal like a startup. We know the rejections, the failures, and the hard-earned lessons that come with building a company. This gives us empathy and understanding for the ups and downs of the founder journey. Read why we started focal here.

To our portfolio, we are high-bandwidth partners who work hard to earn the trust to be the partner you turn to first in the most crucial moments. We work very closely with you as you get your company off the ground. You lead the charge, we challenge and support you to get there faster.

Here's what our founders say about us:

open portfolio.md →

An AI native venture firm

We’re building focal like the companies we back at the cutting edge.

Our firm is a learning system with an autonomous sourcing engine that quantifies every signal we can reach; agent swarms and rankers trained on every decision we make; a self-improving codebase that lets us build our own harness, tailored to our needs, and increase our velocity with each step the frontier does, and MUCH more.

Meaning we treat the firm itself as the model: Every signal is an input, every decision is a label, every outcome is a delayed reward, and the humans are the part of the loop that still sees what the data can't. The firm gets better with every decision, every new data source and every model release, on a self-improving codebase.

In charge of this is Amine who's an exited technical founder with an AI infra background.

01 focalOS

1.4M
nodes in the graph
317k
signals read
2,632
PRs

02 commits

Week of 9 Nov513 PRs
lab
Findings that become fixes: The system now turns its own findings into proposals a partner approves and it then ships.
field
Thesis-driven sourcing experiments: Each investment thesis now runs as a bounded search experiment that updates on partner decisions and founder responses.
lab
Resumable research pipelines: Long-running research now survives outages without losing work it has already paid for.
field
Unified diligence memory: Research, conversations and primary evidence on a company now accumulate in one shared record both partners reason from.
lab
Resilient model routing: Model choice is now a live setting, and work reroutes to another provider when one fails.
field
Deal state as a first-class object: Every live opportunity now carries its evidence, its obligations to the founder and its next decision in one continuously updated state.
lab
Independent evidence capture: Mail and calendar evidence is now captured on its own track and catches itself up after any outage.
field
Research-grade first meetings: Every first conversation now opens with a synthesised research brief and closes with a memo built to be challenged.
lab
Teachable decision agents: Any decision the system makes can now be traced, corrected by a partner, and learned from.
field
Compounding judgement: Each partner decision now trains the firm's models of what matters, so institutional judgement compounds instead of resetting.
lab
Evidence-carrying changes: Every proposed change now arrives with its own checks and a measured estimate of its impact.
field
Context-complete meetings: Every founder meeting is now bound to its full deal history, so the firm's accumulated context arrives before the conversation does.
lab
Unbiased measurement: Accuracy is now scored on a random sample of real decisions, so the system cannot flatter itself.
field
Portfolio intelligence from ground truth: Portfolio reviews are now computed from the firm's ledger of record, so every figure is traceable to its source.
lab
Self-promoting models: A model that beats the incumbent on held-out decisions now goes live on its own.
field
Lossless inbound: Every founder message is now captured, resolved to its company and routed to the right partner, even through upstream outages.
lab
Total evaluation coverage: No model acts inside the firm without a measurement attached to it.
field
Zero-decay founder relationships: Every conversation now resolves into a decision or a commitment, so no founder relationship decays through inattention.
lab
Closed-loop deployment: Every change now ships with a hypothesis and is graded against live production an hour later.
field
Continuous founder discovery: Emerging founders now surface within hours of the first converging signals, each matched to the thesis that predicted them.
lab
Evidence-gated self-correction: The system can revise its own models, but only when the evidence clears a statistical bar.
field
Attention as a measured resource: Every company that reaches the firm is now traced end to end, turning partner attention into something we can measure and optimise.
lab
Self-repairing infrastructure: The system now reads its own production telemetry and writes its own performance fixes.
field
Portfolio construction as a live model: Reserves and follow-on capacity are now planned continuously against the live pipeline, so allocation moves with the opportunity set.

03 infrastructure

graphPostgres property graph, provenance on every edge
identityKeyed resolution, model veto, reversible merges
learningGold labels, 30% hashed holdout, Wilson-gated
inferenceFrontier planner, open-weight executors
judgmentTyped-judgment models, 35 versioned contracts
agents150+ typed MCP tools, audit-or-fail writes
deliverySecret-free merge gate, deploys graded at +60 min
stackPython, TypeScript, Next.js, Supabase, Fly, Vercel
as of 2026-11-18

Your early GTM, supported by 200+ operators

Distribution decides more outcomes than ever.

That's why we spend a lot of our time on your early GTM and plug you into The GTM Circle, our invite-only network of 200+ revenue executives at today's leading Seed through Series D startups.

90% of our founders draw on it regularly.

Members of The GTM Circle include:

  • // Revenue Lead @ Anthropic
  • // Head of Product Marketing @ Anyscale
  • // Partnership Lead @ Langchain
  • // Head of Growth @ Deel
  • // Head of CS @ Notion
  • // Head of Growth @ Delphi
  • // B2B Product Marketing Lead @ Lovable
  • // Head of CS @ Coder
  • // Head of GTM Strategy & Ops @ Harvey
  • // and 200+ more

open gtm-circle.com →

© 2026 focal — hand-built, no ad trackers, screen-reader aware (type 'a11y'). view source; there's a note for you. or: curl focal.vc · working with focal · terms & privacy

Command palette

Jump anywhere