Devin Finzer

A Founders Guide to The Long Build™ - Physical AI Thesis Part III This one is for the builders. In this final post, my partner Will Porteous and I along with insights from builders in the trenches and investors who play the long game share our guide for founders building intelligence in the real world. A few core takeaways: • Sim before steel: Runway is precious, model the system behavior • Capital intensity is not the enemy — misaligned capital is • Build complete solutions before you talk platform • Hardware opens the door, software wins the room • Trust is earned through resilience, not hype

10

1 Comment

💥 Metropolis Technologies Raises $𝟭.𝟲𝗕 💥 Funding comes in 2 rounds: $500M equity round led by LionTree and $1.1B from J.P. Morgan. Metropolis is now 𝘃𝗮𝗹𝘂𝗲𝗱 𝗮𝘁 𝗮𝗿𝗼𝘂𝗻𝗱 $𝟱𝗕. With the fresh funding, the plan is to take their "recognition and payment-automation" tech beyond parking assets. This would include retail stores, drive-throughs, gas stations, hotels and partnerships with real estate/asset owner. Congrats to Alex Israel, Peter Fisher, Courtney Fukuda, Travis Kell, Ryan Hunt,Chris Sherman, Robert Bland, Paul Rooney, Lookman Olusanya and the rest of the Metropolis family. #fundraise #parkinginvestment #parking #parkingindustry #metropolois Parking Today Media Parking Network B.V. National Parking Association International Parking & Mobility Institute (IPMI) Tim Maloney Advisory Tim Maloney Reservations Lab

128

7 Comments

Figma just acquired AI startup Weavy for ~$200M. AI workflows are changing how creative work scales. Here’s why founders should pay attention 👇 Weavy is a Tel Aviv–based AI image & video generation startup. This is Figma’s most important strategic move since Adobe’s attempt to buy Figma was blocked. We’ve entered the AI design consolidation era, where major platforms are becoming AI-first for design and workflow automation: → Canva acquired Affinity (to expand into pro design) → Adobe acquired Rephrase.ai (AI video generation) → Freepik acquired Magnific (AI image upscaling) Because here’s the competitive reality: • Adobe controls the production suite • Canva controls lightweight brand creation • Framer controls design-to-web automation Companies now need to own workflows end-to-end to dominate the market. That means controlling: creation + automation + distribution Because in software, public markets reward leverage. It’s no longer about who has the best design tools but who automates the canvas itself. --- Our team at FE International is more bullish than ever on the rise of AI-first businesses. If you're a founder curious about your company’s true value, get a free valuation. (Link in comments) 👇

34

9 Comments

Fun fact: Brendan, Adni's technical co-founder, was in the second batch of Y Combinator ever. YC S06! He also has a previous successful exit (sold to UnitedHealthcare), patented an AI algorithm before it was cool, and is a published author... NBD! How has the startup ecosystem evolved over the past 20 years? Brendan shares his wisdom below. 💭 #startups #startupfounder #ycombinator #entrepreneur

16

1 Comment

Thrilled to announce that Together Fund is investing in Sentra, alongside a16z speedrun! You track results in Jira. Decisions in Notion. Conversations in Slack. But the reasoning, the debates, trade-offs, and context behind why you chose A over B, disappears into what we call "Dark Matter." A decision made in March looks insane by July because no one remembers the constraints that made it smart. I lived this firsthand at Twitter scaling from 800 to 8,000 employees, and at Google while launching AI Overviews to billions at planet scale. The problem isn't process. Process is compensation for something deeper: organizational amnesia. An organization’s "Systems of Record" doesn’t solve this, they encode it. They store what happened, never why. That's why we are investing in Sentra. Sentra is the always-on collective memory that eliminates organizational amnesia by maintaining accurate context for all members and agents, functioning as an operational nervous system. It connects to every channel where work happens, meetings, Slack, email, code commits, docs, calendars, and treats them not as artifacts to search, but as living signals to synthesize. The fleeting and the permanent, unified into a memory that understands. The founding team is built for this: - Jae Gwan Park (CEO): Product-first founder, memory systems research at UofT and MIT - Ashwin Gopinath (CSO): Former MIT professor, created "Reflexion" (NeurIPS 2023), agents that learn from mistakes, 2x founder - Andrey Starenky (CTO): Early Vapi engineer, ex-IBM, built to process enterprise-scale data firehose Together is an operator-led fund. We invest in problems we've lived. This is one of them. Many congrats Jae, Ashwin and Andrey, we are so excited to partner with you! Read the full thesis: https://lnkd.in/gixj9cE4 Book a demo: https://www.sentra.app/ #OrganizationalMemory #AI #Sentra #TogetherFund #a16z #ContextGraphs

71

3 Comments

This Abram Brown piece captures something I'm seeing everywhere: Tools like Claude Code are giving founders way more agency. For many, it's a reminder of the old days when they were ICs or running tiny companies. To be clear, this is not every founder. But those that were already hands on, in the weeds, in the codebase have just been handed Thor's Hammer.

35

5 Comments

Exciting to see the incredible progress from Peptone, a company we’ve backed since the early days at Walden Catalyst Ventures. Peptone just announced Peptron-O, developed in collaboration with NVIDIA — a groundbreaking AI engine that brings ensemble-first structure prediction to the disordered proteome. This is a major scientific and technological milestone, targeting one of the most complex and previously inaccessible areas of biology: intrinsically disordered proteins (IDPs). By shifting the paradigm from static models to dynamic ensembles, Peptron-O opens new doors for drug discovery — especially in areas like oncology and neurodegeneration, where IDPs play critical roles. We’re proud to have supported Peptone on their mission to build the infrastructure for next-gen molecular physics. Their work is a great example of how deep tech can transform life sciences. Read more about the announcement: 👉 https://lnkd.in/eygnKQ67 #AI #Biotech #DrugDiscovery #DeepTech #LifeSciences #WaldenCatalyst #Peptone #NVIDIA #PeptronO with: Kamil Tamiola, Ph.D. Francis Ho Shankar Chandran Nicolas Autret Andy Kau Roni Hefetz Victoria Slivkoff

56

1 Comment

Researchers at Meta have rethought the decoder for RAG to reduce demands on the key-value cache. REFRAG leverages pre-computed, compressed chunk embeddings as approximate representations, feeding these embeddings into the decoder. Why does this matter? The result is a huge reduction in latency, measured by 30X acceleration in TTFT (Time To First Token) according to the authors. There is lots of detail to unpack in the paper but this suggests a considerable improvement to RAG. One to watch ... See the paper: https://lnkd.in/eg2ymfP7

3

Decision modeling is key. Intent + decision traces created by agents need to be modeled to create self learning or ever improving decision quality ( measured via outcomes), while the criteria is non stop shifting. Process modeling - aka crud on system of records will also get eroded and merged with decisions models being stored. This is where Agebt plane and human plane converge.

6

New Slope Client Work - Brand Identity, Motion, Website Design, Product Icon Library Doma is a purpose-built blockchain to modernize and unlock the liquidity of over 362M existing and future domains. Doma by D3 recently closed a $25M Series A fundraise led by Paradigm, the leading crypto venture firm founded by Coinbase co-founder Fred Ersham and Sequoia Capital Partner Matt Huang. Today is special because the Doma Testnet is officially live 🎉 They are launching with a $1M USDC Developer Fund to support builders shaping the DomainFi economy. --- Our brand identity for Doma positions them as the leader in the Domain industry. Our logomark is a bespoke, geometric symbol that puts Doma at the center of the DomainFi economy. It positions Doma as an architectural foundation of the domain ecosystem that bridges Web2 to Web3. Inspired by the idea of a modular ecosystem, every element of our system - from the logo mark, iconography, to the linear system, were crafted to demonstrate the power of enabling the tokenization of domains, leading to infinite possibilities from fractional ownership, domain-backed lending, and enhanced digital identity and branding. #agency #casestudy #website #crypto #brand

44

7 Comments

"I can't trust an AI black box solution" Hot takes like this are why I love gathering founders and operators together. I recently hosted two dinners at re:Inforce alongside CJ Moses, Paul Vixie, Hart Rossman, Nancy Wang, and Daniel Bartus. Though the groups had different backgrounds, we noticed some common trends: 1: Data: AI is dramatically accelerating the amount of data being generated and moved across services. More data means more surface area of sensitive content. This impacts everything from storage to security, and more. 2: Identity Security: Humans are the weakest link, and AI will increase the number of identities (human, service accounts, etc.) by 100x in coming years. Incumbent providers aren't capable of managing this scale and complexity, and often rely on manual processes to deliver compliance outcomes for employees. 3: Secure AI Enablement: Security practitioners are interested in AI solutions that enable them to maximize impact with fewer resources, but many recognize the risks and challenges in adopting these solutions. They want to trust the vendors, but still verify. Some black box solutions have proven difficult to sell. A huge thank you to the incredible founders, practitioners, and friends of Felicis that came to make the dinners so spectacular! cc Jacob DePriest Alex Bovee Luke O'Malley Tanya Verma Vrajesh B. Eric Tan Prabhath Karanth George Barnes Janusz Wasiolek Marcus Johnston

138

5 Comments

We were recording a podcast episode last week and the question came up: why do studios actually fail? I expected the conversation to land on bad venture selection. That's what most people point to. But three quieter patterns kept surfacing — and they're the ones that actually kill studios. Sequencing failure. Doing the right things in the wrong order. Raising before you have a track record. Hiring a full team before validating a single venture. Everything looks productive — you're shipping, hiring, raising. But the order is wrong, and the compounding works against you instead of for you. Incentive misalignment. The studio operator wants equity upside. The institutional sponsor wants de-risked innovation. The founder wants autonomy. None of those are wrong. They're just different objective functions in the same room. The misalignment doesn't surface in year one. It surfaces the first time someone has to make a hard call — and the people in the room realize they want different outcomes. The governance gap. Studios start lean, which is smart. Then they stay lean on governance, which is fatal. No kill criteria. No stage gates. No framework for when to double down or shut down. Death by accumulated decisions that nobody had the structure to stop. What stuck with me after that conversation: all three of these are invisible from the inside. Bad company selection is loud. These are silent. And by the time you feel them, they've been compounding for months.

14

1 Comment

OpenAI just released HealthBench - an open, physician - validated benchmark for evaluating LLMs in healthcare. Why it matters Existing med-QA datasets don’t mirror the messy, multi-turn reality of clinical work. HealthBench fills that gap with 5,000 multilingual conversations spanning lay-user to clinician scenarios, each graded against a custom rubric written by 262 physicians across 60 countries. The rubric is scored automatically by a GPT-4.1 grader whose agreement with doctors matches inter-physician agreement, giving a scalable yet trustworthy signal . Benchmark design principles * Meaningful – evaluates real-world impact, not board-exam trivia * Trustworthy – physician-created rubrics, physician-validated meta-eval * Unsaturated – plenty of headroom; two variants (Consensus & Hard) deliberately stress today’s best models How current models stack up * Frontier OpenAI models improved 28% on HealthBench in the last few months and now beat Claude 3.7 Sonnet and Gemini 2.5 Pro (Mar’25) * Smaller “nano” models deliver GPT-4o-level performance at 25× lower cost, redefining the cost-performance frontier * In head-to-head tests, April ’25 models (o3, GPT-4.1) produced responses that physicians couldn’t further improve, surpassing both un-assisted and model-assisted physician baselines What is even better - Use & Evolve It * Paper, code, and the full dataset are on GitHub - ideal for benchmarking your own models or fine-tuning domain-specific agents * If you’re building health-AI, start running HealthBench early; tracking worst-case performance (worst-of-n) can expose failure modes before users do HealthBench turns “medical QA” into a rigorous, physician-grounded systems test. Use it to measure progress, surface reliability gaps, and push toward LLMs that clinicians can trust in the wild. https://lnkd.in/d-kcaM_x #AI #ArtificialIntelligence #Healthcare #HealthTech #LLM #HealthBench #startups

4

🌵Latest work for Anima, the biomimicry studio, a venture studio dedicated to nature-inspired innovation. They bring together an experienced team with deep expertise in biomimicry, AI, design, business, and finance to help founders take their ventures from 0 to 1. We partnered with their team to build a unique 3D universe, inspired by organic life and blended with laboratory and AI experimentation. We then designed an immersive website that highlights the art direction while preserving the same atmosphere and vision. Explore their universe → anima.ai Huge congrats to everyone involve Anima team : Sandra REY Alex de las Heras Branding : Clara Hevia Aranguren 3D : Beaucoup x Aurélien Thomas Website : Beaucoup

167

24 Comments

Investors (esp deep tech investors) generally think about things/bet on "layers": the application, the infrastructure, the model. I think that's becoming the wrong way to look at it? The real opportunity is in the feedback loop *between* the layers. A new social app creates demand for new hardware, which in turn demands a new kind of intelligence - I'm calling it "The Agency Loop". My latest newsletter explores this thesis through three signals: the "weaponized transgression" of Cluey, the AI-driven materials science in Nature, and the architectural critiques of Kenneth Stanley. Bet the loop, not the layer :) You can read the full analysis here: https://lnkd.in/ggdVGEda #vibecap #vibecapital #vc #deeptech #AI #cluely

4

1 Comment

🚀 Fundraising update 🚀 We’re building the Modular Super-App Core (MSAC) — patent-pending infrastructure for modular, installable, and monetizable app features. MVP launches this month. We’re raising $750K–$1M pre-seed to reshape how apps get built — and how they earn. Targeting the $400B+ app economy. Infinite scale, built-in. If you invest in platforms, creator economies, or dev tools — or know someone who does — now’s the time. Reach out. Let’s build the future. @PoweredByMSAC on X Your intro could change the future. (video by yours truly 🎬)

3

2 Comments

2 years ago today, A16Z announced they put $100M into Pinecone, as everyone and their mother tripped over each other to invest in vector databases. Given how little standalone vector DBs are valued now, the hype cycle is wild to look back on. Quick timeline below: Late 2022: • ChatGPT launches. • Early RAG architectures emerge. • First buzz around Pinecone, Weaviate, Milvus, Chroma. Q1 2023: • Massive demand surge for vector search. • Vector DBs seen as critical infra. • Heavy VC excitement (“new database category!”). Q2 2023: • Peak hype. • Pinecone raises $100M, Weaviate raises $50M. • Every AI infra diagram has a vector DB layer. Q3 2023: • Saturation hits. • Postgres, Redis, MongoDB add native vector search. • Clouds bundle it. Q4 2023: • Hype fades. • Vector search becomes a commodity. Early 2024: • Market tightens. • Attention shifts to agents, fine-tuning, orchestration.

763

117 Comments