Dylan Reid

What Our Top Performing Investments at SpringTide Ventures Have in Common Our most transformative healthcare startups, including PathologyWatch, OpenLoop, IgniteData, and Tava Health, all share three traits in common: 1. Cheetah CEOs These CEOs are exceptional at: ✅ Speed without chaos. They move fast—but never at the expense of execution. ✅ Hiring with precision. They don’t just hire for skills; they hire for speed and alignment. ✅ Betting bold. They go where legacy players have overcomplicated things for decades—and they simplify. They don’t wait. They don’t overthink. They act and embrace the process of failing quickly, often, and cheaply. They learn from these failures to iterate at rapid speed and quickly orient their way to the north star of success more nimbly than others. 2. No Unpaid Pilots The single biggest trap in healthcare venturing may well be the unpaid pilot. It’s a slow death. Months—sometimes years—of proving your value with nothing to show for it. The best founders? They refuse. They find a way to generate real revenue immediately. They structure deals that prove their worth—without giving away their product for free. They don’t waste time on handshakes that lead nowhere. 3. Integrated Solutions Win Point solutions don’t win in healthcare anymore. The best startups don’t just build products. They build systems. They understand the bigger picture—the way everything connects. They integrate seamlessly into existing workflows instead of fighting them. They don’t just add value—they make their solution the only viable choice. This integration often comes with lower margins in the early years of finding and building out product Market fit, but we gladly accept lower margins for such ambitious visions. Healthcare is brutal. Most startups don’t make it. But the bestredefine their categories. They’re led by Cheetah CEOs like Daniel Lambert, Jon Lensing, Zach Taft, MBA, and Dallen Allred—founders who execute relentlessly, think in integrated solutions, and get to real revenue fast.

109

7 Comments

With RL techniques like GRPO, it’s become easy to fine-tune small models to outperform foundation models on vertical tasks. We’re open sourcing Osmosis-Apply-1.7B: a small model that merges code (similar to Cursor’s instant apply) better than foundation models. Links in comments! Using foundation models for high specificity, low complexity tasks like code merging can be overkill - instead, it’s better to use a specialized model instead for higher performance and latency. We fine-tuned the model on 100K real-world code commits where the model needs to merge a code edit snippet into an original code snippet. On the test dataset (subset of CommitPackFT), foundation model performance ranged between 0.77-0.93 reward score. In comparison, Osmosis-Apply-1.7B achieved a 0.98 reward score - while also being 3X-5X more affordable than the next cheaper model AND ~10X faster! Why this matters: the best AI agents use multiple models to ensure task specialization and reliability at scale - on top of cost and latency benefits as well. Links to the model and full write-up (with reward function, hyperparameters, etc.) are in the comments - reach out if you’re interested in using reinforcement learning!

239

18 Comments