Writing/Product/ Introducing RELAI: Verifiable Continual
Product

Introducing RELAI: Verifiable Continual Learning for AI Agents Backed by $6.9M in Funding

June 10, 2026 — Today, we’re launching RELAI, a verifiable continual learning platform for AI agents, and announcing $6.9 million in total funding to scale our mission. The funding includes a newly secured $5.4 million pre-seed round led by .406 Ventures with participation from AITFund and other strategic investors, along with $1.5 million in prior investment support from Non sibi Ventures and TEDCO.

Verifiable Continual Learning Is the Real Frontier

As enterprises move AI agents into production, keeping them reliable after deployment has become one of the hardest unsolved problems in AI. Agents fail unpredictably, and fixes often create silent regressions, leaving teams stuck in a cycle of prompt patches, rerun evaluations, and reactive debugging.

The issue is not that agents cannot learn. A growing set of methods already update prompts, memory, skills, and models from experience. The issue is that almost none of that learning is verified against what already works. An update that fixes today’s failure can silently break behavior that worked yesterday, and no one notices until it reaches production.

We built RELAI to solve this problem.

RELAI turns failures, traces, evaluations, and human feedback into replayable learning environments. Instead of treating a failure as an isolated bug, RELAI converts it into a reusable learning signal that can be optimized against and verified. Every failure becomes an opportunity for durable improvement.

What makes RELAI different is how improvement is validated. Most systems check for regressions after a change is made. RELAI keeps regression control inside the optimization loop itself. Every proposed improvement is continuously validated against a growing portfolio of prior learning environments while it is being searched, not after it is shipped.

Just as importantly, RELAI identifies where a repair belongs. A failure might require a prompt change, a tool wrapper, a memory update, a workflow adjustment, a model-routing decision, or a code-level repair. RELAI diagnoses the root cause and applies the smallest durable change at the layer where it belongs, instead of piling every fix into an ever-growing prompt.

In early deployments, this approach increased a financial services agent’s validation score from 39% to 80% and improved a healthcare proof-of-concept from 62% to 96%, without the manual debugging loops those gains would normally require.

We believe the next frontier of AI is not simply making agents more capable. It is enabling them to learn continuously from real-world experience without becoming more fragile. As agents take on increasingly important workflows, organizations will need systems that transform experience into reliable, generalizable improvement.

That is the future we are building toward at RELAI.

We’re opening limited-access onboarding today ahead of our broader public release. If you’re building or deploying AI agents and want to help shape the future of continual learning, we’d love to hear from you.

Join the waitlist at relai.ai.

LP
Laurent Py
CEO, RELAI
CEO of RELAI, serial entrepreneur

Keep reading.

All posts →
Continual Learning for AI Agents: How to Actually Build It cover
ResearchContinual Learning for AI Agents: How to Actually Build ItSoheil Feizi · 2026-05-28 · 9 min
Get the next post.
One email when something new ships from RELAI. Research write-ups, product launches, and field notes on building AI you can trust.