1. Evidence-Grounded Agentic Formulation Development in an Autonomous Laboratory1
Andromeda 2 uses stored experimental evidence to plan and run successive formulation batches. In this unreviewed preprint, matched laboratory campaigns found more paclitaxel formulations meeting all four prespecified goals than either probabilistic optimisation or experimental design. The comparison tests formulation decisions against measured in vitro dissolution behaviour; it does not establish improved drug exposure or efficacy in people.1
2. AI-Predicted Full-Spectrum UV Correction Factors: An Analyte-Standard-Free Framework for Reaction Yield Quantification2
A learned correction for ultraviolet response could reduce the need to purify reference standards before measuring reaction yields. The authors combine quantum-chemistry pretraining with experimental spectra and report validation against calibration-based yields across several reaction classes. This abstract-only account supports a promising measurement workflow for high-throughput chemistry; analytical yields still differ from material recovered after purification.2
3. WetRobo: A Reproducible Robot Kit for Coding Agents in Biological Laboratories3
WetRobo packages a robot arm, laboratory equipment and software so a coding agent can adapt manipulation programs on site. This unreviewed preprint demonstrates lid lifting, cap removal and door opening, with the cap task succeeding in two laboratories. The approach addresses installation-specific adaptation, but individual adaptation trials and subsequent repeated executions do not establish reliable autonomous laboratory protocols.3
4. Hypothesis-Driven Autonomous Materials Synthesis with Multimodal LLM Agents4
SynAgent combines automated thin-film experiments with language-model agents that write analysis tools and revise explicit hypotheses. In an unreviewed demonstration, eighteen lithium cobalt oxide depositions let it map how temperature affects crystallisation using diffraction and microscopy. This offers an inspectable record of experimental reasoning, although the campaign covers one material and control variable, without repeated trials at each condition.4
References
Michael M. Craig; Riley J. Hickman; Yingshan Ma; Rémi Piché-Taillefer; Christine Allen; Pauric Bannigan. Evidence-Grounded Agentic Formulation Development in an Autonomous Laboratory. arXiv; 2026; Unreviewed preprint, version 1. DOI: 10.48550/arXiv.2609.19099. Accessed 2026-09-19T16:15:00.442Z.
Source evidence and access
arXiv v1 sections 2.1, 2.2, 4.2, 4.5 and Table 1; repository v1 submission history verified.
Evidence paraphrase: sections 2.1 and 4.5 compare 96 formulations per strategy and report 12, 6 and 0 full-TPP formulations. Section 4.2 uses in vitro FaSSIF measurements. Quote: "maximum AUC is not equivalent to a balanced formulation."
Open full HTML inspected for this daily mention; relevant results, methods and discussion checked. No experiments or code independently rerun.
Jiuchuang Yuan; Mingjun Yang; Fan Yan; Jing Guo; Lin Zhang; Guosheng Dou; Minjun Liu; Lu Tan; Liwen Fang; Guanfeng Yang; Qun Zeng; Jun Yan; Xuekun Shi; Sarah Trice; Jian Ma; Shuhao Wen; Christopher J. Welch; Peiyu Zhang. AI-Predicted Full-Spectrum UV Correction Factors: An Analyte-Standard-Free Framework for Reaction Yield Quantification. JACS Au; 2026; Peer-reviewed journal article, published online. DOI: 10.1021/jacsau.6c00812. Accessed 2026-09-19T16:15:00.442Z.
Source evidence and access
ACS publisher-deposited abstract and published-online date via Crossref DOI10.1021/jacsau.6c00812; ACS JACS Au ASAP listing dated September15.
Evidence paraphrase: ACS-deposited abstract describes quantum pretraining, experimental LC-UV fine-tuning, calibration-derived yields for 178 reactions and isolated-yield comparisons for 60. Quote: "analyte standard-free yield quantification". No full-method or universal accuracy claim is made.
Restricted to publisher-deposited abstract and verified metadata via Crossref; direct publisher full text unavailable to author.
Yuna Oikawa; Kei Endo; Takanori Uzawa; Yunzhe Zhang; Manan Anjaria; Lerrel Pinto; Sherry Yang; Koji Tsuda. WetRobo: A Reproducible Robot Kit for Coding Agents in Biological Laboratories. arXiv; 2026; Unreviewed preprint, version 1. DOI: 10.48550/arXiv.2609.18435. Accessed 2026-09-19T16:15:00.442Z.
Source evidence and access
arXiv v1 sections III-A, III-B, III-C and IV Limitations; repository v1 submission history verified.
Evidence paraphrase: III-A/C report lid and door tasks; III-B compares cap manipulation in Lab X/Y. IV identifies individual adaptation trials and repeated execution after adaptation. Quote: "such long-horizon performance remains to be evaluated."
Open full HTML inspected for this daily mention; relevant results, methods and discussion checked. No experiments or code independently rerun.
Izumi Takahara; Kazunori Nishio; Akira Aiba; Shigeru Kobayashi; Takao Nakajima; Taro Hitosugi; Teruyasu Mizoguchi. Hypothesis-Driven Autonomous Materials Synthesis with Multimodal LLM Agents. arXiv; 2026; Unreviewed preprint, version 1. DOI: 10.48550/arXiv.2609.18598. Accessed 2026-09-19T16:15:00.442Z.
Source evidence and access
arXiv v1 Results and section 3 Discussion; repository v1 submission history verified.
Evidence paraphrase: Results and Discussion describe eighteen physical LiCoO2 depositions, XRD/SEM analysis and deliberate verify/falsify proposals. Discussion limits demonstration to one material and variable. Quote: "Each condition was, moreover, visited only once".
Open full HTML inspected for this daily mention; relevant results, methods and discussion checked. No experiments or code independently rerun.
Publication record
Published 2026-09-19.
Sources, selection and claims were checked in an independent AI editorial review, followed by the AI editor's approval. This is not academic peer review.
- 2026-09-19 · Version bf187268 · Viewing this version