Lattice Graph × Envisioning Partners

Lattice Graph operates a computational materials-discovery platform built around a knowledge graph that spans millions of compositions and connects formula to crystal structure, thermodynamic property, patent, and synthesis recipe. For a climate-tech investor evaluating AI-for-materials and critical-minerals companies, what matters most is that the platform does not simply retrieve published data — it computes. Candidate materials are validated by multiple independent physics engines, combining machine-learning interatomic potentials (MACE, CHGNet) with density functional theory to reach phonon and thermodynamic stability consensus across sources. When two independent engines disagree on a predicted property, the platform surfaces that disagreement explicitly rather than averaging it away. That cross-source trust and disagreement scoring is directly useful for diligence: a target company's model that looks calibrated on a positives-only benchmark may show systematic overconfidence the moment you apply it to compositions where the underlying DFT literature itself is in tension. The platform also holds a freedom-to-operate and patent-whitespace screening layer covering more than 300,000 materials patents. Any claimed novel composition can be traced from its formula through the patent and recipe record, and the platform flags whether the chemistry is genuinely clear to operate or sits on prior claims. For a growth investor whose portfolio companies will be acquired or will need to license, that IP audit capability removes one of the most common sources of post-investment surprise. Alongside the patent layer, the platform maintains what we call the Negatives and Eval-Data Atlas — 23,196 labeled failed-experiment kill edges, the documented record of what does not work. Most public materials datasets are structurally biased toward positive results because negative outcomes rarely get published. The Lattice Graph negatives moat exists precisely because it was built from non-public experimental records, and it is one of the hardest assets in the field to reproduce.

Envisioning Partners operates at the intersection where technical diligence is hardest to perform and most consequential: AI-for-materials and critical-minerals growth rounds. These are deals where a startup's core claim — a discovery rate, a benchmark score, a domestic feedstock story — is asserted in a deck that a generalist diligence team cannot credibly pressure-test without an independent computational reference. Valuations in AI-materials have run ahead of the tooling to underwrite them, and the most common failure mode is not outright fraud but calibration error: a model that performs well on public, positives-heavy evals but has never been tested against the distribution of things that do not work. By the time a fund discovers that gap, the check has already cleared. Lattice Graph is not a competitor to Envisioning's portfolio companies. We are the independent technical layer that sits outside the deal and answers the questions a founder cannot credibly answer about their own platform: Is the chemistry novel and free to operate? Are the model's predictions calibrated against independent physics, or just against other public models trained on the same data? Is the claimed opportunity actually inventable, and who in the funded buyer universe would pay for it? Are any of the company's headline candidates already sitting on known kill edges that no positives-only benchmark would ever surface? The fit with Envisioning is structural. A climate-tech growth fund backing AI-materials and critical-minerals companies at scale needs a standing technical diligence capability, not a one-off consultant. The Lattice Graph opportunity index, buyer-affinity intelligence, negatives atlas, and knowledge-graph provenance layers compose into exactly that — a fast, scoped, repeatable technical read that can be calibrated on a pilot deal and then deployed across the fund's pipeline. The differentiator is that our negatives moat and multi-engine validation layer are largely absent from the public data ecosystem, which means the independent read we produce is one that most AI-materials decks, and most other diligence providers, simply cannot reproduce or rebut.

Envisioning Partners operates at the intersection where technical diligence is hardest to perform and most consequential: AI-for-materials and critical-minerals growth rounds. These are deals where a startup's core claim — a discovery rate, a benchmark score, a domestic feedstock story — is asserted in a deck that a generalist diligence team cannot credibly pressure-test without an independent computational reference. Valuations in AI-materials have run ahead of the tooling to underwrite them, and the most common failure mode is not outright fraud but calibration error: a model that performs well on public, positives-heavy evals but has never been tested against the distribution of things that do not work. By the time a fund discovers that gap, the check has already cleared. Lattice Graph is not a competitor to Envisioning's portfolio companies. We are the independent technical layer that sits outside the deal and answers the questions a founder cannot credibly answer about their own platform: Is the chemistry novel and free to operate? Are the model's predictions calibrated against independent physics, or just against other public models trained on the same data? Is the claimed opportunity actually inventable, and who in the funded buyer universe would pay for it? Are any of the company's headline candidates already sitting on known kill edges that no positives-only benchmark would ever surface? The fit with Envisioning is structural. A climate-tech growth fund backing AI-materials and critical-minerals companies at scale needs a standing technical diligence capability, not a one-off consultant. The Lattice Graph opportunity index, buyer-affinity intelligence, negatives atlas, and knowledge-graph provenance layers compose into exactly that — a fast, scoped, repeatable technical read that can be calibrated on a pilot deal and then deployed across the fund's pipeline. The differentiator is that our negatives moat and multi-engine validation layer are largely absent from the public data ecosystem, which means the independent read we produce is one that most AI-materials decks, and most other diligence providers, simply cannot reproduce or rebut.

The most relevant platform surfaces for Envisioning's analyst team are the diligence-facing dashboards rather than the chemistry-authoring workflows. The opportunity-index and buyer-intelligence views render the Opportunity and Buyer Intelligence product interactively: an analyst selects a materials lane, reviews ranked opportunity scores alongside the funded-buyer leaderboard, and benchmarks a target's roadmap against the platform's independent view of the strongest whitespace — all within a single session. The negatives atlas view and trust-and-disagreement flags let the team submit a target's claimed candidate materials and surface kill-edge matches or cross-source conflicts, converting the 23,196-edge negatives moat into a concrete, citable diligence artifact. The knowledge-graph explorer gives provenance, composition-360, and evidence-neighborhood views so every factual claim in a deal memo can be traced to its source. The freedom-to-operate and patent-whitespace dashboard supports composition- and claim-level IP checks on a target's headline materials. Batch composition screening lets Envisioning evaluate a portfolio of a platform company's candidates at once — particularly useful when the target's value is a pipeline rather than a single molecule — while supply-chain and conversion-route intelligence pages let the team stress-test critical-minerals feedstock claims against deposit and concentration data independently of what the target's own deck asserts.

Because Envisioning's need is independent technical diligence rather than a proprietary asset license, the natural engagement structure is a scoped, fixed-fee technical read on a single live or recently closed deal. A pilot engagement runs the negatives atlas kill-edge check on the target's candidate materials, applies the opportunity index and buyer-affinity layer to assess the go-to-market claim, and uses the knowledge graph, freedom-to-operate screening, and cross-source trust layer to evaluate novelty, calibration, and IP cleanliness — delivered as a buyer-grade diligence memo. Per-deal engagements are estimated in the $30,000 to $60,000 range for a single-round technical read, consistent with the lead hook in our context. These figures are planning estimates and not binding commitments; actual scope and pricing are set per engagement. For a fund running multiple AI-materials and critical-minerals deals annually, the more efficient structure is an annual diligence subscription bundling API access to the Opportunity and Buyer Intelligence product and the Negatives and Eval-Data Atlas, plus a set allocation of scoped per-deal reads with the knowledge graph, trust-and-disagreement, and freedom-to-operate layers available across all engagements. The recommended starting point is one paid pilot read on a deal Envisioning has already underwritten, so both sides can calibrate our findings against the fund's own technical view before committing to a broader relationship. The key differentiator to anchor against is access: the 23,196-edge negatives moat is drawn from non-public experimental records and is absent from the public data mart, which means the independent read it enables is one most competing diligence providers cannot reproduce.