Catherine Havasi

Systems first. AI next. The order matters, a lot. Over the past few weeks, Evan Allen and I have been digging into agentic workflows - reading papers, testing tools, and mapping patterns across different implementations. One insight keeps standing out. Agentic workflows aren’t inherently “AI.” They’re decomposition systems. The core idea is simple: take a complex task, break it into smaller, well-defined steps, execute those steps independently, then reassemble the output.  Who performs each step is secondary. 👉 Some steps are best handled by machines: speed, recall, pattern matching. 👉 Others still benefit from humans: judgment, accountability. 💡 Once you see it this way, the false binary between “automated” and "manual” disappears. For human-in-the-loop systems - like what we're building with Nexus - it shifts the goal from replacing people to assigning work to whoever (or whatever) does it best, step by step. AI becomes a delegate, not an endpoint. That also forces more discipline. You have to be explicit about what the task actually is, what good looks like, and where judgment truly matters. Sloppy workflows don’t magically improve when you add AI - they just fail faster. I suspect the most durable systems won’t be the most automated ones, but the most intentional about where humans stay in the loop. Where have you seen the balance work well, or break down?

14

2 Comments

Rough news on #kuzu being archived - startups are hard and Semih Salihoğlu + Prashanth Rao did so much in ways I value, and the same architectural principles we've been quietly tackling in GFQL. For those left in the lurch for an embeddable compute-tier solution to graphs, #GFQL should be pretty fascinating yet also familiar (ex: Apache Arrow-native graph queries for modern OSS ecosystems), and hopefully less stress due to a sustainable governance model. Likewise, as an oss deeptech community, we add interesting new bits like the optional record-breaking GPU mode with NVIDIA #RAPIDSAI . If you're new to it and seeing this: #GFQL, the graph dataframe-native query language, is increasingly how Graphistry, Inc. and our community work with graphs at the compute tier. Whether the data comes from a tabular ETL pipeline, a file, SQL, nosql, or a graph storage DB, GFQL makes it easy to do on-the-fly graph transforms and queries at the compute tier at sub-second speeds for graphs anywhere from 100 edges to 1,000,000,000 . Currently, we support arrow/pandas, and arrow / #nvidia #RAPIDS as the main engine modes. While we're not marketing it much yet, GFQL is already used daily by every single Graphistry user behind-the-scenes, and directly by analysts & developers at banks, startups, etc around the world. We built it because we needed an OSS compute-tier graph solution for working with modern data systems that separate storage from compute. Likewise, data is a team sport, so it is used by folks on teams who have to rapidly wrangle graphs, whether for analysis, data science, ETL, visualization, or AI. Imagine an ETL pipeline or notebook flow or web app where data comes from files, elastic search, databricks, and neo4j, and you need to do more on-the-fly graph stuff with it. We started building what became GFQL *before* Kuzu because it solves real architectural & graph productivity problems that have been challenging our team, our users, and the broader graph community for years now. Likewise, by going dataframe-native & GPU-mode from day 1, it's now a large part of how we approach GPU graph deep tech investments throughout our stack, and means it's a sustainably funded system. We are looking at bigger R&D and commercial support contracts with organizations needing to do subsecond billion+-scale with us so we can build even more, faster (hit me up if that's you!), but overall, most of our users are just like ourselves, and the day-to-day is wanting an easy OSS way to wrangle graphs in our apps & notebooks. As we continue to smooth it out (ex: we'll be adding a familiar Cypher syntax), we'll be writing about it a lot more. 4 links below: ReadTheDocs, pip install, SOTA GPU benchmarks, and original aha moment + Russell Jurney Ben Lorica 罗瑞卡 Taurean Dyer Bradley Rees

25

2 Comments

I recently wrote about my experience serving as General Chair for the first NeurIPS satellite conference in Mexico City—what worked, what we learned, and what it revealed about expanding access to #AI research globally. This was a meaningful experiment in how we convene as a field, and I hope these reflections are useful as the community continues to think about inclusion, scale, and collaboration in AI.

22

5 Comments

Very proud to announce Buoyant Ventures latest investment in Hammerhead to help unlock more power for AI and data center growth. My partner Laura Katzman (Dyer) led this deal and our research into efficient computing which helped us win this competitive and over-subscribed deal. Thanks to Maeve Allsup for your excellent explanation of the opportunity. https://lnkd.in/gSM4PadB

64

2 Comments

MIT FutureTech now has a substack: "Missing the MIT FutureTech Conference? Read our reflections on the conference via Substack, and explore recorded presentations now available on Youtube Mixture of Experts’ new Substack highlights a selection of ideas presenters raised and connections we found interesting, loosely organized around three themes: 1) Technical and economic constraints to AI-fueled explosive growth, 2) AI adoption and the labor market, and 3) AI automation in science. YouTube links to the talks are included throughout the post, see comments below for links" Great work from Zachary Brown Alexander F. Sebastian Sartor and Lucy Yan

40

4 Comments

Very excited for this announcement! Many of the world’s hardest and most valuable problems live in scientific and engineering domains where answers must obey physical laws, design constraints, and provable correctness. That is exactly the challenge the team at Axiomatic is taking on. And there is acute opportunity in semiconductors and photonics. Device physics follows well-established mathematical formalisms, fabrication happens in highly controlled fabs and foundries, and the design space, while complex, is governed by clear physical rules. That combination of strong theory, controlled experimentation, and well-defined parameter spaces makes these fields some of the most promising places for AI to have immediate impact. If AI systems can reason over those constraints and verify correctness, the implications for generative design are enormous. From photonic devices to next-generation semiconductor architectures, AI could begin exploring design spaces for chips that power computing, communications, sensing, and so much more. Axiomatic’s work is a key step toward making AI truly useful inside real scientific and engineering workflows. Huge congratulations to Dirk Englund, Marin Soljacic, Frank Koppens, Joyce Poon, Jake Taylor and the entire Axiomatic team. Excited to see what happens next.

36