Catherine Havasi

Data in a silo on its own has no use. But giving people the tools to contextualize and form opinions on the data is what will set medical providers apart. One of the biggest societal need areas for data interoperability is within healthcare, but we also see the same need and opportunity set within other essential sectors like manufacturing and mobility. Loved chatting with First Builder and data interoperability extraordinaire Jason Prestinario about this.

9

This is a great article on why small language models outperform large language models in many use cases. Faster, cheaper AI is ultimately the outcome from SLMs, because LLM's can burn through thousands of times as many tokens to complete a task as they burn through them in an internal monologue they use to reason through an answer.  SLM's are focused on solving very specific problems and do it much faster using far less power.  https://lnkd.in/gH4UF-gB

23

LLMs are getting commoditized. The pace of core model breakthroughs has clearly slowed. We are no longer seeing step-function jumps from “just a bigger model”. For most real systems today: • architecture matters more than the base LLM, • planning, orchestration, memory, and verification dominate quality, • the model itself is increasingly a replaceable component. In my own code factory, I can swap GPT, Claude, or Gemini with minimal impact. There are differences in latency and edge cases, but they no longer define the system. The next real breakthroughs will not come from GPT-6. They will come from new architectures built around models, not inside them. LLMs are becoming infrastructure. Systems are becoming the product.

1

One of our latest companies DN Capital - directionally this is where agentic AI can help augment humans in the workplace rather than replace humans. Congratulations to our partner Guy Ward Thomas and the rest of the DN Capital AI investment team on this newest portfolio company. Nenad Marovac Thomas Rubens Ian Marsh Gülsah Wilke Raoul Oscar F. Jack Pierrepont Maddie Coughlin Celine Xiao

19

At Logical Health, we’re combining AI with declarative, logic-based programming to deliver highly accurate results in healthcare. Super cool to see MIT exploring a similar approach: their new paper Teaching LLMs to Plan shows how step-by-step logical reasoning boosted planning accuracy from 28% to 94%. The future of AI isn’t just bigger models. It’s smarter, more reliable ones that align with real-world needs. Arxiv Link: https://lnkd.in/gnN-XYGs

72

7 Comments

Great lineup of events in Massachusetts coming from the Massachusetts AI Coalition over the next few weeks. Why do IRL events matter? Three reasons: 1.) You'll get more practice building with AI and learn about AI, 2.) You'll meet all of the great founders & operators in Boston at their companies in case you decide you want to work at them, 3.) You'll make friends. A community of friends is a powerful community. A community of zoom aquaintances is no community at all. By meeting people in real life, you have the opportunity to connect on what each other are building, yes, but you'll also learn about each other's personal lives... hobbies... interests... families. You will make friends. It sounds cheesy, but to build a vibrant community, you have to care about the people beside you. Put in the work to get out into the community and the friends will come. Request to join an event. Your older self will thank you for it. We'll see you out there my friends. Nice work Sonika Vuyyuru for taking lead on these posts, as well as to Sam Iannone, Michelle Fitzgerald, Beth York, Liam Saven for producing content regularly for the AI Coalition. Well done!

22

4 Comments

Jotted down some end of year biopharma reflections while waiting to board a flight. Likely no surprises. 2025 was tough, but there's a hint of currents shifting for the better 🤞 1. BIG gulf between what people think scientists are clambering for, when it comes to AI tools, and what scientists are *actually* asking about, using regularly, and willing to pay for. "LLM for this, agent for that" really misses the deeper-rooted challenges and bottlenecks. 2. Lots of killing of discovery, shifting emphasis to later stage assets. Even saw this at orgs where the discovery engine is working (producing clinic stage assets that continue to get good clinical readouts). I understand this reactivity to markets and investor sentiment, but it’s nevertheless sad to see strong scientific engines be shut off, teams laid off, and novel discovery stopped as a result. TBD what longer term effects will be over next several years. 3. It often takes the experience of having gone through a cycle to realize what problems you want to avoid at all costs. Our most well-aligned and motivated champions have been people who directly experienced the painful alternatives ('no action' or 'build-it-yourself'). Conversely, those who haven't had to grapple with the problems before often maintain a "we can just do everything ourselves" stance. 4. A lot of work is being outsourced. When managing these CRO relationships, complexity can balloon quickly. Our partners have increasingly turned to Kaleidoscope.bio to drastically streamline this pain. 5. There is painful disconnect between how much time people waste on preventable stuff, and how much time/budget/awareness leadership will provide to address this. I encourage leaders to empower their team to solve problems that will help them move faster, even if they as a senior leader may not deal with the day-to-day (and thus may not feel it directly). 6. Seems to be an increasing number of scientific PMs spearheading operations (we at Kaleidoscope like this).

58

7 Comments

mpathic is becoming THE partner to the AI industry as they work to safeguard users from adverse mental health and safety outcomes As millions of users — including large numbers of young people — increasingly turn to AI chatbots as their first-line “counselors” and confidants, Seattle-based startup mpathic is stepping in to ensure those digital agents don’t provide dangerous advice when it matters most.

33

4 Comments

New work on Rethinking Thinking Tokens: LLMs as Improvement Operators: https://lnkd.in/errhNuuz Reasoning training encourages LLMs to produce long chains of thought (CoT), improving accuracy via self-checking but increasing context length, compute cost, and latency. This work studies whether frontier models can achieve better trade-offs, higher accuracy with lower cost. The paper develops a simple yet effective Parallel-Distill-Refine (PDR) procedure: Generate diverse drafts in parallel, Distill them into a compact textual workspace, and Refine conditioned on this workspace. This decouples context length from total token count, allowing control over compute via parallelism. PDR yields higher accuracy than long CoT at lower latency. Training an 8B model with RL to align with PDR further shifts the Pareto frontier. On math benchmarks, PDR achieves +11% (AIME 2024) and +9% (AIME 2025) over single-pass baselines. With Lovish Madaan, Aniket Didolkar, Suchin Gururangan, John Quan, Ruan Silva, Manzil Zaheer, Sanjeev Arora, and Anirudh Goyal.

78

1 Comment