Agon Avdimetaj

How AI Makes Medicine More Humane: An Analysis by Chatfuel CEO A new article from Fedor Pak, CEO of our portfolio company Chatfuel, has been published. Pavel wrote a column in MedCity News about how AI is transforming patient interactions in clinics. Key points from the article worth taking note of: ▪️ AI shouldn't replace doctors; it should assist. Medical clinics don't need an "all-knowing digital doctor." A much more useful tool is a "polite assistant with maximum functionality," who reminds patients about appointments 24/7, forwards instructions, answers common questions, and helps with scheduling. ▪️ Time savings and revenue growth. In real-world cases, communication automation allows clinics to process 3-5 times more requests without hiring new staff, reduces the number of patients who don't show up for appointments without warning by up to 45%, and increases the volume of paid procedures by 20-40% per week. ▪️ Automation makes clinics more humane. Doctors spend an average of almost 6 hours maintaining electronic medical records for every 8 hours they spend seeing patients. When AI takes over routine tasks, doctors arrive at their appointments less burned out and more focused. ▪️ It's not just about chatbots; communication automation is just the first step. AI support for clinical decision-making, predictive follow-ups, automated documentation, and more innovative anamnesis collection processes will follow. If you work in HealthTech or clinics, I recommend reading the entire article. It's an efficient and insightful look at the implementation of AI in healthcare. 🔗 Read the article on MedCity News: https://lnkd.in/eW8Nr-s6

19

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!

240

18 Comments

How AI is changing venture building + numbers & best practices of PSL startup studio ($50M AUM studio funds + $180M AUM venture funds): 1) 3 dedicated studio funds: $12M (2015) → $16M (2018) → $20M (2023) for a 25-person build team (salary, sprints, infrastructure). Venture arm: $80M (2018) + $100M (2021) — invests in external startups and PSL studio spin-outs (but never prices later rounds). Geographically focused on the Pacific Northwest. 2) 5-stage company creation funnel: – Ideation (problem hunt, AI research),~4,000 ideas – Validation (customer calls, landing page), 470 validated ideas – Creation (MVP, sales), 50 created products – Spin-out (legal, PSL seed round, CEO equity), 41 spin-outs – Scale-up (external VC funding, recruiting). 37+ scale-ups with external funding. 3) Throughput: 4-6 new spin-outs per year target. Typical time of a startup in the studio: 3–4 months. Fastest build ever: 6 weeks. At any moment: ~2 projects in Creation, 3-5 in Validation. 4) How AI is shrinking Ideation & Validation phases with 2 PSL internally created tools: – Spark – researches the market, auto-drafts decks & landing pages, creates an initial ICP list, then A/B-emails prospects until it gets 10 yes/no signals. Founders start with data, not vibes. – Mesh – maps 70,000+ operator/investor contacts of the PSL team, scores fit, then sends emails and tracks reply rates; the warmest path shows up as a Slack thread the same day. Together they compress “customer proof + talent hunt” to < 24 h instead of 2-3 weeks earlier. 5) Seed checks. Non-AI spin-out: $300-500K from PSL Ventures. AI spin-out: $1.5M co-led ($750K PSL Ventures + $750K AI Studio Fund with Mayfield). Follow-on rounds are syndicated — outside VCs set price; PSL follows its pro-rata. 6) Typical equity split is 50/50 between studio & founding CEO. But can vary a bit based on circumstance (founder track record, founder-idea fit narrative, existing work done by the founder prior to the Studio engagement...) 7) The 4 founder archetypes PSL is partnering with: – Successful repeat founder (typically, 40-50 y.o.) They value the speed & the focus, which is possible due to the studio team + higher conviction in an idea you will bet 5+ years of your life on. – Product Manager to CEO (~25-40 y.o.) They value studio expertise and support as a first-time CEO. They can also find some strong ideas inside the studio if they don't have one. – Young Tech Builder (YTB) < ~30 y. — typically miss and undervalue a new company's business development/sales roles. Studio helps with it. – Corporate exec/domain expert. They know how to get the maximum from the corporate partner but need support as a first-time CEO. Watch the full interview with T. A. McCann – Managing Director @ Pioneer Square Labs to learn insights about venture building with AI and managing a startup studio: youtu.be/YP-i5mhXNiY Bonus from PSL: Would you like to see a demo video of how Spark works? I can DM the link if interested

105

10 Comments

The Apex Burn Protocol: Alkoin Deflationary Architecture The Apex Burn Protocol outlines a strategic financial framework designed to ensure the permanent deflation of the Alkoin cryptocurrency. By integrating digital finance with tangible utility, the system utilizes remittance fees and real estate rental yields to fund automated open market buybacks. These purchased tokens are permanently destroyed, effectively shrinking the total supply as the ecosystem's commercial activity expands. Additionally, the protocol implements a transaction tax and provides incentives for long-term holders to further reduce the number of tokens available for sale. This multi-layered approach aims to create consistent upward price pressure by linking token value directly to real-world economic growth. Through this "machine," the platform aligns international money transfers with a mathematical model for investor profit.

1

$110M raised in 5 months.   A one-year-old startup just raised $85M in a Series A round led by Redpoint Ventures, with participation from 8VC, Shopify Ventures, Autotech Ventures, and others.   That tells us investors believe:   – Workflow-native AI (trained on logistics data + processes) creates defensible moats.   – Adoption curves are faster when AI feels like a teammate, not another SaaS login.   – The TAM is huge. Logistics is a trillion-dollar industry with margin pressure and labor bottlenecks.   Logistics is just one example. Similar vertical plays are emerging in:   Legal → Contract review agents. Finance → AI underwriters, wealth management agents. Construction → AI safety monitors and project schedulers. Healthcare → Ambient clinical documentation, revenue cycle AI.   Augment’s raise is a datapoint in the larger trend:   📌 Capital is moving from broad AI platforms into applied, industry-specific AI.   What AI are you building?   _______________ P.S. My newsletter (Join 20K+ subscribers): https://lnkd.in/dNVbGxXv My reading list: Reading List | https://lnkd.in/dS8XJ9_W

105

9 Comments

CVCs, if they are well designed and well run, can offer startups value-add in addition to traditional VCs, such as built-in infrastructure for scale. GV (Google Ventures) leads with an impressive 90 unicorn investments, followed by Salesforce Ventures (42) and Intel Capital (40). Central Eurasia needs more Corporate Venture Capital, as it can generate significant benefits both for corporations themselves and for the startup ecosystem as a whole ! Future Laboratory Ventures

15