Early-stage VC firm based in Silicon Valley. Investing in the next wave of unicorns building GenAI infrastructure.

Joined November 2024
8 Photos and videos
Pinned Tweet
Truly inspired by everyone at the @agihouse_org Hackathon "Think Slow and Think Deep w o1". Shoutout to @HelenSandHill for sharing the necessity of breaking the compute wall, the model wall, and the data wall to reduce hallucinations and enhance data accuracy! #tech #Hackathon
1
62
Proud of @TurionSpace for landing a spot on the Top 50 Dynamism List! Congrats on pushing boundaries and Unleashing the Potential of Earth Orbit #FoundersX #DynamismUnleashed @HelenSandHill
11 Mar 2025
BOOM! Turion just landed on the Top 50 Dynamism List – proof we’re shaking up the game! Join the revolution, follow us, and let's Unleash the Dynamism and Accelerate our Galactic Destiny! #DynamismUnleashed
1
43
Grok3 drops: grok.com/?referrer=website. DeepSearch deep thinking tech—set to disrupt research & multiply productivity. #AI #Grok #LLM #FoundersX
35
🚀 The Evolution of AI Reasoning: Are we approaching #AGI? The frontier of #AI isn’t just about making models bigger—it’s about making them think better.
1
8
DeepSeek R1 proved reasoning can emerge purely through reinforcement learning (RL)—no supervised fine-tuning. The model teaches itself how to think by being rewarded strictly for correctness.
1
16
💡 Key Insight: Reasoning can emerge without massive supervised #CoT data, making training cheaper & more efficient. ✔️ Lower data requirements → More cost-effective #AI development ✔️ Data-driven models (O-series) may still outperform RL-based reasoning in unstructured tasks
13
DeepSeek-R1 is redefining AI economics, proving that reasoning models don’t need billion-dollar compute budgets. By optimizing efficiency instead of brute-force scaling, this open-source breakthrough offers a new blueprint for high-performance AI with lower resource constraints.
1
30
🔥 Performance vs. OpenAI o1 📌 Math (AIME 2024): 79.8% pass@1 (vs. OpenAI-o1’s 79.2%) 📌 Coding (Codeforces Elo): 2029 Elo (96.3% percentile, comparable to OpenAI-o1) 📌 MATH-500: 97.3% pass@1 (vs. 96.4%)
1
19
💡 Why This Matters ✅ High-performance reasoning with smaller compute budgets ✅ Open-source AI—accessible to all ✅ Software-driven efficiency—optimized for existing hardware #AI #DeepSeek #MachineLearning #LLM #OpenSource #FoundersX @HelenSandHill
20
MIT’s 3D Chip Breakthrough: A Game-Changer for AI Engineers have cracked the code for multilayered chips, overcoming traditional limitations of horizontal design. Using a “growth-based method” and nucleation techniques, this innovation could revolutionize computing power for AI
1
20
Small reasoning models can be powerful too: DeepSeek-R1-Zero is trained with large-scale reinforcement learning (RL) without supervised fine-tuning (SFT), showcasing impressive reasoning capabilities #VCinsights #AIModels #Technology #DeepLearning
20 Jan 2025
🚀 DeepSeek-R1 is here! ⚡ Performance on par with OpenAI-o1 📖 Fully open-source model & technical report 🏆 MIT licensed: Distill & commercialize freely! 🌐 Website & API are live now! Try DeepThink at chat.deepseek.com today! 🐋 1/n
79
🎉 Incredible night at the pre-#JPM25 dinner in SF with the visionary founders of Omniscope & 1910 Genetics! 🚀 We loved the cutting-edge #BioTech advancements & the amazing progress from these #startups Good luck to all founders at #JPM25 #FoundersX #Startups @HelenSandHill
114
Incredible storytelling meets cutting-edge creativity! 🎥✨ DreamFlare AI’s latest collaboration, The First Artist, with Emmy-winning producer @ericrharcher, is a journey through time—exploring the origins of creativity 35,000 years ago. #FoundersX #Entertainment #GenAI
16
Happy New Years from the team at FoundersX Venture!
1
9
Highlights from #NeurIPS: 1️⃣ Scaling Together: AI and systems must co-evolve. 2️⃣ End of Pre-Training: the next frontier—#Superintelligence 3️⃣ Do All Models Converge?: As AI scales, do all models eventually represent the same “world”? #AI #SystemsAI #NeurIPS2024
1
37
Key Takeaways from #IEDM: 1️⃣ Scaling below 5nm is hitting limits as silicon struggles with electron scattering. 2️⃣ The future lies in 2D materials: New growth methods are enabling single-crystalline, low-temp, high-yield integration with CMOS. #Semiconductors #TechInnovation
1
14
Day 11: OpenAI now integrates seamlessly with coding apps like PyCharm and popular note-taking tools like Notion and Apple Notes. Harness the power of o1 models directly within the apps you use every day. #OpenAI #ProductivityTools #TechInnovation
19 Dec 2024
Day 11: A new way to work with ChatGPT openai.com/12-days/?day=11
1
66
Day 10: 📞 Introducing 1-800-ChatGPT—your personal AI assistant, just a call away! ✅ Need quick answers on the go? ✅ Brainstorming ideas while driving? ✅ Hands-free help for busy schedules? Call now and experience the future of personal assistance! #AIInnovation #ChatGPT
18 Dec 2024
1
29