Two months since the April update. This one is less about a feature list and more about a shift in how we work — and in what Athanor is becoming.
1. 100+ Conversations
Over the last months we talked to more than 100 people interested in the product — founders, solo GPs, venture club operators, accelerator teams, and acquirers. Most of them received demo access to the platform and ran it on their own deals and ideas, not on our curated examples.
That matters more than any feature we could have shipped in the same time. A demo on someone else's real deal flow is brutal: it surfaces the confusing form field, the score that needs explaining, the report section nobody reads. Much of what shipped this quarter — and most of what got fixed — traces directly back to one of those sessions.
It also sharpened our focus. The plan for the next stretch is deliberately simple:
- Serve the clients we have. Pilots and active users get priority — their feedback loop is the product roadmap now.
- Find more like them. If you run deal flow — as a fund, a club, an accelerator, or a corporate team — and want to see the platform on your own pipeline, reach out. Demo access is how most of our current clients started.
2. GemScore V5 — Evaluations, Rebuilt as Packs
The biggest engineering effort of the quarter is GemScore V5 — a ground-up rebuild of the evaluation pipeline on a pack-native architecture, now rolling out as the default for VC startup evaluation.
What changes for you:
- Audience-aware reports. The same evaluation renders as a founder report (what to improve and why) or an investor tear-sheet (decision-grade synthesis, IC-ready) — assembled from the same underlying data, not two separate documents drifting apart.
- An evidence chain under every claim. Each conclusion in the report traces back to its sources. Sources are trust-weighted: independently verified metrics carry the most weight, data-room documents are weighed by provenance, and self-reported claims are clearly marked as such. Agents see the trust level — and reason accordingly.
- Verify, dispute, re-run. If a finding is wrong, dispute it with evidence. The pipeline re-runs the affected analysis instead of charging you for a full re-evaluation.
- Hardened scoring. Real pilot deals taught us where automated scoring fails. V5's deterministic rubric layer now uses calibrated caps and explicit flags instead of blunt automatic zeroes, and fraud detection distinguishes a confirmed finding from a hypothetical risk mentioned in passing.
- Export-grade PDFs. Cover page, firm branding, deal team, confidentiality settings — reports you can put in front of an IC without reformatting.
V5 has been running with partner pilots on live deal flow for weeks. The calibration work above is the direct result.
3. The Athanor SDK — Same Shape, New Verticals
V5's real significance is underneath the VC pack. Rebuilding the pipeline forced us to extract what every serious evaluation has in common — and turn it into infrastructure. That infrastructure is the Athanor SDK.
Every pipeline on the platform is now a pack, built from four parts:
- A manifest — a YAML file declaring the pack's name, domain, version, agents, grading scale, and resource budgets. The platform loads it without code changes.
- Agents — each implements a single protocol and carries the actual domain expertise: what to look at, what good looks like, what kills a deal.
- Deterministic stages — debate, rubric scoring, synthesis. Agents disagree; the stages resolve that disagreement the same way every time, so two runs on the same evidence don't produce two different verdicts.
- An evidence chain — every claim in the output traces back to a source, trust-weighted, inspectable.
What the SDK actually brings is a separation of concerns. Domain experts define what to evaluate and how to judge it. The platform supplies everything an evaluation product needs to run in the real world: authentication, retries, evidence aggregation, scoring, audience-aware report rendering, PDF export, versioning. A new vertical is a manifest and a set of agents — not a new product, not a new codebase.
The catalog today
Three packs already run on the SDK — browse them at /pipelines:
- VC Startup Evaluation — the GemScore pipeline you know. Five specialised agents (Team, Market, Business, Product, Risk), adversarial debate, and a rubric layer with hard kills and soft penalties. Outputs Potential and Readiness on a 0–100 scale, plus financials, sensitivity analysis, and SWOT.
- Roblox Acquisition — acquisition-focused evaluation of Roblox UGC studios. Fourteen agents covering audience, retention, revenue, discovery, IP chain, and clone risk, synthesised by a Deal Architect into a recommended offer with tiered earnout templates and holdback structure.
- M&A Due Diligence (early preview) — pre-LOI screening of acquisition targets: governance, seller credibility, and financial health, condensed into a buy/wait/walk recommendation.
Notice what these three share: almost nothing — except the shape. A startup, a game studio, an acquisition target. Different inputs, different experts, different scoring. Same manifest, same stages, same evidence chain, same report engine.
Build your own vertical
That's the invitation. If you hold deep domain knowledge and need to assess anything — deals, projects, companies, assets, applications — reach out, and we'll build your vertical together. You bring the expertise: what matters in your domain, what the red flags are, how a seasoned practitioner judges quality. We turn it into a pack that evaluates at scale, with the evidence discipline built in.
Write to [email protected]. The conversations we're having now decide which verticals come next.
Where We Are
The 100+ conversations changed how we describe Athanor to ourselves. It's no longer just a startup-evaluation tool — it's becoming evaluation infrastructure: one platform, pluggable pipelines, evidence-first reports, for any deal that needs a decision.
We're still in Beta, still iterating fast. But the loop is working: real users, real deals, real corrections — every release sharper than the last.
Join us — your first Screening is free, every month. And if you want a guided demo on your own pipeline, say hello.
Questions? Hit us on Discord or contact support.
