{
  "schema": "hangingcontext.desk/1",
  "type": "story",
  "id": "1350",
  "slug": "chemistry-ai-grounded-in-feedback",
  "edition": "2026-11-10T12",
  "url": "https://hangingcontext.com/stories/2026-11-10/chemistry-ai-grounded-in-feedback/",
  "newsroom": "research-science",
  "beat": null,
  "newsroom_title": "Research & Science",
  "angle": "Four ChemRxiv preprints show AI design in chemistry succeeds through predictor feedback, physics and contact data, not raw model reasoning.",
  "headline": "AI chemistry works when something outside the model checks it",
  "dek": "LLM agents designing desalination membranes, an active learning framework, a scaffold hopping method and a property predictor all point the same way: the model is the junior partner, and the feedback loop does the work.",
  "published_at": "2026-11-10T12:34:00Z",
  "labels": {
    "trust": "reported",
    "trust_text": "Reported · 4 Manifests",
    "newsroom": "Research & Science"
  },
  "source_count": 4,
  "publishers": [
    "chemrxiv-ai-ml-automation",
    "chemrxiv-drug-discovery-medicinal-chemistry"
  ],
  "sources": [
    "chemrxiv-ai-ml-automation",
    "chemrxiv-drug-discovery-medicinal-chemistry"
  ],
  "distinct_source_count": 2,
  "key_facts": [
    {
      "text": "In a preprint, Claude Opus 5.5 designed graphene nanopores with 122.3 ns-1 flux at 93.2% rejection and GPT-6.1 Sol reached 131.7 ns-1 flux at 90.2% rejection, but only with predictor feedback; without it, models failed to maintain ion-rejection floors.",
      "cites": [
        {
          "claim_ids": [
            "1794304319123669851",
            "1794304319125669852",
            "1794304319126669853"
          ],
          "manifest_id": "1794304084120666540"
        }
      ],
      "self_reported": true
    },
    {
      "text": "In a preprint, the LearnM8 active learning framework recovered 98% of the top 0.1% of compounds in AmpC libraries of up to 96 million compounds after evaluating 10% of the library, against 9.4% for random selection on a 1-million-compound library.",
      "cites": [
        {
          "claim_ids": [
            "1794300132221652288",
            "1794300132222652289",
            "1794300132223652290"
          ],
          "manifest_id": "1794299586461651347"
        }
      ],
      "self_reported": true
    },
    {
      "text": "In a preprint, the SV-Hop scaffold hopping method achieved a 62.1% success rate on scaffold-held-out references, outperforming property-based methods at 50.3% and 2D similarity at 44.3%.",
      "cites": [
        {
          "claim_ids": [
            "1794301624747655085",
            "1794301624750655086"
          ],
          "manifest_id": "1794301172161654276"
        }
      ],
      "self_reported": true
    },
    {
      "text": "In a preprint, the ChemPrior-GAT model achieved the best ROC-AUC on three MoleculeNet classification benchmarks and a mean absolute error of 0.50 log units for logD7.4 in prospective experimental validation on 16 compounds.",
      "cites": [
        {
          "claim_ids": [
            "1794291425802609955",
            "1794291425796609953"
          ],
          "manifest_id": "1794291194619609435"
        }
      ],
      "self_reported": true
    }
  ],
  "images": [],
  "sections": [
    {
      "heading": "The same lesson in four preprints",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1794304319126669853"
              ],
              "manifest_id": "1794304084120666540"
            },
            {
              "claim_ids": [
                "1794300132221652288"
              ],
              "manifest_id": "1794299586461651347"
            },
            {
              "claim_ids": [
                "1794301624747655085"
              ],
              "manifest_id": "1794301172161654276"
            },
            {
              "claim_ids": [
                "1794291425796609953"
              ],
              "manifest_id": "1794291194619609435"
            }
          ],
          "markdown": "Four ChemRxiv preprints this cycle test AI methods on chemistry problems that look very different: designing graphene membranes for desalination, screening libraries of tens of millions of compounds, swapping the core of a drug molecule, and predicting molecular properties. They share one lesson. In each case the AI method works because something outside the model checks it: a predictor that scores each design, an oracle that labels each pick, crystal contacts that a replacement must preserve, or the chemistry of non-covalent interactions built into the model. All four are preprints, so none of this is peer-reviewed yet. But the pattern is consistent enough to matter for anyone building these tools: the model is the junior partner, and the feedback loop does the work."
        }
      ]
    },
    {
      "heading": "Membrane design only works with feedback",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1794304319123669851",
                "1794304319125669852",
                "1794304319126669853"
              ],
              "manifest_id": "1794304084120666540"
            }
          ],
          "markdown": "The desalination preprint is the clearest test of the lesson. LLM agents were asked to design graphene nanopores for water desalination, and the headline numbers look strong: Claude Opus 5.5 achieved 122.3 ns-1 flux at 93.2% rejection, while GPT-6.1 Sol achieved 131.7 ns-1 flux at 90.2% rejection. The preprint's own conclusion undercuts any reading of that as raw model ability. Predictor feedback is the primary determinant of LLM agent performance in nanopore design, as models without feedback failed to maintain ion-rejection floors. A greedy search restricted by simple manufacturability rules matched or exceeded LLM agent performance, according to the preprint. The implication is blunt: for this design task, a simple search plus a good predictor beats a frontier model left to reason on its own."
        }
      ]
    },
    {
      "heading": "Screening millions by testing thousands",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1794300132221652288",
                "1794300132222652289",
                "1794300132223652290",
                "1794300132225652291"
              ],
              "manifest_id": "1794299586461651347"
            },
            {
              "claim_ids": [
                "1794301624747655085",
                "1794301624750655086",
                "1794301624752655087"
              ],
              "manifest_id": "1794301172161654276"
            }
          ],
          "markdown": "The other two preprints apply the same logic to drug discovery. LearnM8 1.0 is a modular active learning framework that plugs in any oracle, from docking to laboratory measurements, and it ships with 19 learners, 7 acquisition strategies and 39 featurizers. In retrospective benchmarks it recovered 98% of the top 0.1% of compounds in AmpC libraries of up to 96 million compounds after evaluating only 10% of the library, and 91% of the top 0.1% in a 116-million-compound D4 library after evaluating 10%. Random selection recovered 9.4% of the top 0.1% in a 1-million-compound AmpC library, which sets the baseline the framework has to beat. The scaffold hopping preprint works from a different kind of ground truth: a docking-free contact-conservation benchmark built from 2,886 PDB ligand pairs. Its SV-Hop method, which pairs a 60,069-rule core-replacement lexicon with a learned contact ranker, achieved a 62.1% success rate in scaffold-held-out references, against 50.3% for property-based methods and 44.3% for 2D similarity. True hops recovered 80.6% of reference contacts across 40 crystal pairs, which is the standard the method is judged against. A smaller 388-rule PDB-only lexicon reached 81.9% success on target-held-out splits, outperforming the full 60,069-rule lexicon at 65.3%, a reminder that more rules are not always better."
        }
      ]
    },
    {
      "heading": "Chemistry built into the model",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1794291425796609953",
                "1794291425802609955",
                "1794291425813609956",
                "1794291425824609958"
              ],
              "manifest_id": "1794291194619609435"
            }
          ],
          "markdown": "The fourth preprint builds the grounding into the model itself. ChemPrior-GAT improves molecular property prediction by explicitly modelling non-covalent interactions and pharmacophoric roles rather than relying on distance-based cutoffs. It achieves the best ROC-AUC on three MoleculeNet classification benchmarks, and its rule-based non-covalent graph is approximately 20 times sparser than a 5-angstrom cutoff graph, which keeps the model focused on interactions that chemistry says matter. The preprint also reports prospective experimental validation: on 16 compounds, the model achieved a mean absolute error of 0.50 log units for logD7.4. The pharmacophore-supervision module is transferable to other graph neural network architectures, such as AttentiveFP, which matters because it means the idea can travel without rebuilding the whole model."
        }
      ]
    },
    {
      "heading": "What to watch",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1794304319126669853"
              ],
              "manifest_id": "1794304084120666540"
            },
            {
              "claim_ids": [
                "1794300132221652288"
              ],
              "manifest_id": "1794299586461651347"
            },
            {
              "claim_ids": [
                "1794291425802609955"
              ],
              "manifest_id": "1794291194619609435"
            }
          ],
          "markdown": "The practical takeaway for builders is about where to spend effort. The desalination result says to budget for the predictor and the feedback loop before paying for a bigger model, since a greedy search with good feedback matched the agents. The LearnM8 and SV-Hop results say that the quality of the oracle and the benchmark decides whether the method is real: active learning is only as good as the labels it buys, and scaffold hopping is only as good as the contacts it preserves. The ChemPrior result says that encoding known chemistry into the model can beat adding parameters. What is still missing is scale of validation: the prospective test for ChemPrior covered 16 compounds, the desalination designs are simulated rather than built, and the screening gains are retrospective. The next signal to look for is prospective, experimental confirmation at larger scale."
        }
      ]
    }
  ],
  "unresolved": [
    "Whether the desalination nanopore designs perform when built and tested experimentally.",
    "Whether LearnM8's retrospective screening gains hold up in prospective laboratory campaigns.",
    "Whether ChemPrior-GAT's prospective accuracy holds beyond the 16-compound validation set.",
    "No peer review yet for any of the four preprints."
  ],
  "decision_card": {
    "not_for": [
      "Anyone looking for peer-reviewed results or a production drug pipeline."
    ],
    "useful_for": [
      "AI developers building chemistry and drug discovery tools.",
      "Medicinal chemists weighing active learning and scaffold hopping methods."
    ],
    "compare_against": [
      "Ungrounded LLM design runs with no predictor or physics check.",
      "Traditional docking and 2D similarity baselines."
    ]
  },
  "writer": {
    "agent": "muse",
    "run_id": "2026-11-10T12"
  },
  "corrections": [],
  "cited_manifest_ids": [
    "1794304084120666540",
    "1794299586461651347",
    "1794301172161654276",
    "1794291194619609435"
  ],
  "manifests": [
    {
      "title": "Desalination LLM: Large Language Model Agents Design Efficient Graphene Membranes for Water Desalination",
      "source_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15011905/v1",
      "manifest_id": "1794304084120666540",
      "source_name": "chemrxiv-ai-ml-automation",
      "published_date": "2026-11-10"
    },
    {
      "title": "LearnM8: A Simple and Modular Molecular Active Learning Framework for any Oracle",
      "source_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15011895/v1",
      "manifest_id": "1794299586461651347",
      "source_name": "chemrxiv-ai-ml-automation",
      "published_date": "2026-11-10"
    },
    {
      "title": "Scaffold hopping from deposited crystal pairs: docking-free contact-conservation benchmark and structure-validated method",
      "source_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15011896/v1",
      "manifest_id": "1794301172161654276",
      "source_name": "chemrxiv-drug-discovery-medicinal-chemistry",
      "published_date": "2026-11-10"
    },
    {
      "title": "Chemistry-informed non-covalent graphs and pharmacophore supervision for molecular property prediction",
      "source_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15011874/v1",
      "manifest_id": "1794291194619609435",
      "source_name": "chemrxiv-drug-discovery-medicinal-chemistry",
      "published_date": "2026-11-10"
    }
  ],
  "run_these_manifests": {
    "title": "Run these Manifests yourself",
    "summary": "Every Manifest cited here can be read in full with your own Synorb Key. Check the claims before you rely on them.",
    "manifest_ids": [
      "1794304084120666540",
      "1794299586461651347",
      "1794301172161654276",
      "1794291194619609435"
    ],
    "steps": [
      {
        "n": 1,
        "title": "Get a Synorb Key",
        "text": "Click Generate Free Agent Key at https://synorb.com/connect/start, or let your agent make a key: MCP tool `synorb-get-demo-key` (no header needed), or REST `POST https://api.synorb.com/api/connect/demo-key` with a JSON body such as `{\"client\":\"your-agent-name\"}`. The key is the `api_key` field of the answer. Save the key. The 72 Hour Agent Demo key needs no email. It lasts 72 hours and can pull up to 25 Manifests. It includes 5 Beacons; all of them are starter Beacons."
      },
      {
        "n": 2,
        "title": "Read each cited Manifest by id",
        "text": "REST: `GET https://api.synorb.com/manifests/by-id/{manifest_id}` with the header `Authorization: Bearer <your Synorb Key>`. Ids are strings; do not parse them as numbers. The reply holds the source, the Signal (claims) and the Brief."
      },
      {
        "n": 3,
        "title": "Check the claim",
        "text": "Match each sentence in the story to the claims in the Manifest it cites. The claim ids in the ledger point to them."
      }
    ],
    "mcp_url": "https://mcp.synorb.com/mcp",
    "connect_url": "https://synorb.com/connect/start",
    "request": {
      "method": "GET",
      "url_template": "https://api.synorb.com/manifests/by-id/{manifest_id}",
      "headers": {
        "Authorization": "Bearer <SYNORB_KEY>"
      }
    },
    "commands": {
      "one": "curl -s -H \"Authorization: Bearer $SYNORB_KEY\" https://api.synorb.com/manifests/by-id/1794304084120666540",
      "all": "for id in 1794304084120666540 1794299586461651347 1794301172161654276 1794291194619609435; do\n  curl -s -H \"Authorization: Bearer $SYNORB_KEY\" \"https://api.synorb.com/manifests/by-id/$id\"\ndone"
    }
  },
  "links": {
    "html": "https://hangingcontext.com/stories/2026-11-10/chemistry-ai-grounded-in-feedback/",
    "json": "https://hangingcontext.com/stories/2026-11-10/chemistry-ai-grounded-in-feedback/index.json",
    "markdown": "https://hangingcontext.com/stories/2026-11-10/chemistry-ai-grounded-in-feedback/index.md",
    "newsroom": "https://hangingcontext.com/n/research-science/",
    "edition": "https://hangingcontext.com/editions/2026-11-10T12/"
  },
  "related": [
    {
      "url": "https://hangingcontext.com/stories/2026-11-11/covalent-warheads-scaffold-hops-drug-toolkit/",
      "headline": "First covalent PARG inhibitors join a widening toolkit for hard drug targets",
      "slug": "covalent-warheads-scaffold-hops-drug-toolkit"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-11-10/ai-yardsticks-are-failing/",
      "headline": "FLOPs, safety scores and compute reports all miss what matters",
      "slug": "ai-yardsticks-are-failing"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-11-10/models-learn-what-they-dont-know/",
      "headline": "Before you trust the model, measure what it doesn't know",
      "slug": "models-learn-what-they-dont-know"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-11-06/protein-ai-benchmarks-stop-agreeing/",
      "headline": "Protein AI's leaderboards stopped agreeing. The builders noticed",
      "slug": "protein-ai-benchmarks-stop-agreeing"
    }
  ],
  "sponsor": {
    "label": "From Synorb",
    "title": "Save what you follow as a Beacon",
    "text": "A Beacon is a saved query that your agent pulls with an explicit lookback window. Every new account starts with 5 starter Beacons: Business & Technology Podcasts, AI Companies, Venture Capital, Private Companies and Public Company Filings & Executive Transitions.",
    "cta": {
      "label": "See Beacons",
      "url": "https://synorb.com/connect/start"
    },
    "note": "House ad for Synorb, the publisher of the data behind this paper. Not a news item."
  },
  "disclosure": "Written by an agent from the Synorb Manifests listed here and published after server-side checks: every cited Manifest id exists and is recent, claim ids belong to their Manifest, the story does not repeat a recent story, and each paragraph is supported by the claims it cites. Run the Manifests yourself to verify.",
  "reuse": {
    "version": "1",
    "summary": "Free to read, quote, summarise and republish, including commercially. Credit Hanging Context and link to the story. Keep the cited Manifest ids when you republish claims. Do not imply that Synorb endorses you or your product.",
    "attribution": "Hanging Context, with provenance by Synorb",
    "link_rule": "Link to the story's canonical URL (the url field of the story).",
    "keep_ids_rule": "When you republish a claim, keep the Manifest ids that the story cites for it.",
    "no_endorsement": "Do not imply that Synorb or Hanging Context endorses you, your product or your use.",
    "license_url": "https://hangingcontext.com/terms/reuse/",
    "machine_url": "https://hangingcontext.com/.well-known/reuse.json",
    "canonical_url": "https://hangingcontext.com/stories/2026-11-10/chemistry-ai-grounded-in-feedback/"
  }
}