Shelved Drug Patents Are a Goldmine – Speed Is Everything

01 /The Concept

Convexia aims to bring new drugs to market ten times faster than the current industry standard – by building what it calls "the world's first AI-maximalist pharmaceutical company," one in which nearly all the work is done by AI. Here's the pipeline.

The pipeline begins with continuous scanning of pharmaceutical research databases and patent filings for "shelved" drug formulas – compounds that labs discovered but lack the resources to commercialize. Every candidate runs through roughly fifty specialized AI models evaluating efficacy and adverse effects, producing a shortlist of compounds worth pursuing; human specialists then verify those AI assessments and narrow the field further.

From there, a commercial evaluation AI runs demand projections, competitive landscape analysis, and financial modeling on the surviving candidates. A risk-assessment AI simulates the clinical trials that would be required across a range of scenarios before any real-world commitments are made. For each compound that clears all of these gates, an AI generates a pitch deck tailored to specific pharmaceutical manufacturers – explaining why this compound fits their target patient population, synergizes with their existing portfolio, and so on.

The final judgment – scientific rigor, commercial viability, regulatory compliance, clinical risk, and market timing – is handled by human domain experts.

Convexia runs with a minimal internal headcount. For human oversight of AI outputs, it draws on external specialists rather than a large in-house team. The same logic applies to laboratories: it doesn't own any, outsourcing all required research to contract labs. Everything that can be outsourced, is – with one exception: its AI models, which are the company's sole core asset.

The long-term business model is to acquire the compounds that survive the full analysis, conduct clinical trials directly, and then sell successfully validated compounds to pharmaceutical manufacturers who will bring them to market under their own brands.

In the near term, Convexia is also open to licensing individual platform modules to pharma companies for their own use, and to running co-pilots with them to validate pieces of the workflow.

Convexia is currently in the Y Combinator batch and posted on the YC site only a few days ago.

02 /Why It Matters

The most important thing about Convexia's approach is what it doesn't do: it doesn't try to invent new drugs. It finds already-discovered formulas and evaluates their market readiness – then prepares to buy and commercialize the best of them.

This immediately recalls another self-described "AI-maximalist" startup, Rocketable ([covered here](/review/v-obshhem-sluchae-jeto-poka-fantastika-a-v-chastnom-vozmozhnost-na-milliard)), which also graduated from Y Combinator earlier this year and raised $6.5M.

Rocketable wants to build an "AI-maximalist software holding company." Its plan is to acquire SaaS products generating between $500K and $2M in ARR, with at least two years of operating history and a clear path to profitability – and then hand those products over to AI for autonomous improvement. Not revolutionary new features, but continuous incremental enhancement of existing functionality.

That constraint is the whole point: adding genuinely new capabilities to software products with just a prompt or two "remains a dream" in Rocketable's own words. And asking AI to invent new products from scratch is even further from reality. Which is why Rocketable acquires already-validated products rather than trying to ideate them.

Convexia operates on exactly the same principle – applied to pharmaceuticals. AI can't reliably invent new drugs. But it can evaluate thousands of existing formulas and prioritize the most commercially viable ones for investment and commercialization.

03 /Opportunities

Using AI to write copy, draft emails, draw up contracts, build websites, or generate code – even if remarkable – is now ordinary. Nobody calls that maximalism.

Handing a SaaS product to AI for autonomous management, or using AI to source, evaluate, and commercialize drugs – that still sounds like science fiction. Or maximalism.

But Y Combinator has now backed two companies operating in exactly this mode. That's a reasonable signal that what seems impossible today is about to become possible. And the biggest returns will go to the startups already building in that direction.

So: in what other domains could someone build an "AI-maximalist" company – one where AI does the maximum fraction of all the real work?

Don't constrain the thinking with "AI can't do that yet." The technology is moving fast enough that today's limitations are tomorrow's baseline. The goal is to arrive at the right position before everyone else does.