Rishi Taparia (@taps) header
Rishi Taparia (@taps) profile picture

Co-founder at @GarudaVC and host of the @brickxbrickpod. Fully leaned in to suburban dad life.

San Francisco, CA
Joined April 2009
Photo by Rishi Taparia (@taps)Photo by Rishi Taparia (@taps)Photo by Rishi Taparia (@taps)Photo by Rishi Taparia (@taps)Photo by Rishi Taparia (@taps)Photo by Rishi Taparia (@taps)Photo by Rishi Taparia (@taps)Photo by Rishi Taparia (@taps)Photo by Rishi Taparia (@taps)Photo by Rishi Taparia (@taps)
Rishi Taparia retweeted
Our paper is on the cover of Advanced Science this month, so I wanted to share a bit more of the story behind it. I spent years in cognitive neuroscience running large scale human studies which often required thousands of participant hours, months of analysis and many rounds of revisions before arriving at the conclusion or insights.  So we asked a simple question: can an AI system do the whole scientific loop? Can it come up with a hypothesis, test it against real human data, analyse the results, and write it up to a standard that survives peer review? So we tried. We found that the AI systems we built can successfully run scientific experiments end-to-end. Scientific advancement and the methods in which we create knowledge and understanding is dramatically changing. There is a lot I can say about the achievement, but perhaps most important is the development of a system founded on the principles that make science trustworthy. The system doesn’t tell you what you want to hear, it doesn’t change hypotheses or tests to fit the data, and won’t hallucinate an answer it doesn’t have. It writes a pre-registration of the design plan before it sees any data, reports null results, focuses on effect sizes rather than chasing p-values, and checks every citation against the record. Our goal is to advance science and accelerate new discoveries in a rigorous and robust way society has trusted for decades, while freeing researchers to spend their time on the ideas that matter. We believe science and technology will work to better understand the universe, and we intend to accelerate the frontier.  Huge thanks to our brilliant co-authors and collaborators at @ExScienceAI, The University of Queensland, Queensland Brain Institute - QBI, and the University of Sydney. @VentureBeat story here venturebeat.com/ascend-conte… paper here advanced.onlinelibrary.wiley…
5
4
11
1,494
Rishi Taparia retweeted
Today, we're releasing @hardwarebench, the first hardware engineering benchmark built in partnership with the world's most advanced hardware companies. Measurement has always been core to our work at @arenaphysica: the very first version of Atlas shipped with our internal framework for evaluating frontier models on real hardware engineering tasks. HW-Bench evaluates frontier model performance on production hardware data and captures the full hardware context across tools and file formats. Each task is graded against rubrics and golden answers reviewed by subject matter experts at our partner companies, who build some of the most advanced hardware in the world, from the chips behind modern supercomputing to the autonomous systems and vessels redefining defense at sea and beyond (more coming soon). We'll be building towards jobs that span disciplines, like designing, manufacturing, and testing a 140 GHz, 512-element D-band phased array. To start, see how 15 frontier models performed at designing a 2.45 GHz antenna at hw-bench.com.
16
27
10
215
13,745
Rishi Taparia retweeted
Foundry Robotics CEO @adarshkulkarni explains why robots that work 99% of the time still aren't reliable enough to build satellites: "Something that's 90% reliable fails one out of 10 times. Something that is 99% reliable fails one out of 100 times. Something that is 99.99% reliable fails one out of every 10,000 times." "So the difference between 90% and 99.99% is 1,000 times fewer errors. That’s why it’s so hard." "When you're doing manufacturing for critical industries, when you're building drones or batteries or satellites, you really have to be pretty good. You gotta have the nines." @FoundryRobotics
Today, @FoundryRobotics is announcing our $100M Series A, led by @MarlinspikeVC with continued support from @khoslaventures. Together we’re securing America's ability to defend, compete, and innovate by building critical hardware at home with our new 130,000-square-foot factory in Emeryville, California. We’re already shipping to paying customers across energy, telecom, autonomous vehicles, and defense. The next era of American manufacturing starts here. Time to build.
3
6
1
42
6,958
Rishi Taparia retweeted
Today, @FoundryRobotics is announcing our $100M Series A, led by @MarlinspikeVC with continued support from @khoslaventures. Together we’re securing America's ability to defend, compete, and innovate by building critical hardware at home with our new 130,000-square-foot factory in Emeryville, California. We’re already shipping to paying customers across energy, telecom, autonomous vehicles, and defense. The next era of American manufacturing starts here. Time to build.
117
71
69
858
159,141
Rishi Taparia retweeted
It's only Thursday (on a short week), and it's already been a week for the books here at Garuda Ventures and for our LPs. While others wait with baited breath debating the timing and details of the Anthropic IPO at VC dinners, we are over here printing DPI for our LPs - via two acquisitions by Tier 1 enterprise software leaders - and writing new first check investments into explosive markets: 💰 A sizable acquisition of C1.ai by @Workday, bringing the leading agentic identity platform for people and agents to the enterprise agentic system of record for HR, finance, and IT. As Workday put it: "Today's companies run on a combined workforce of people and AI agents". C1 + Workday will secure that hybrid workforce. Congrats to the incredible Alex Bovee and Paul Querna, and the entire C1rew! 💰 Acquisition of our portfolio company ComplianceCow by @ServiceNow. ComplianceCow built agentic continuous controls monitoring - validating controls and exposures in real time for cyber risk and compliance - and will nicely complement and strengthen ServiceNow's Integrated Risk Management (IRM) product line. Congrats to the amazing Raj Krishnamurthy and the killer CCow team! 🚀 Committed to lead a pre-seed round for a robotics simulation infrastructure company. This market is taking off and we're pumped to back a hungry and technically ambitious team at Day 1 that is playing to win. More to come on this one soon! :) Thank you as always to the LPs and founders who bet on us and what we are building. You are how and why we do what we do, and Rishi and I are grateful for your partnership. Back to work!
4
1
2
16
1,081
Rishi Taparia retweeted
first autonomous AI research with real human participants! congrats to @ExScienceAI on their peer-reviewed paper, also on the cover of Wiley's Advanced Science: advanced.onlinelibrary.wiley… total runtime: ~17 hours agents involved: 50+ tokens used: 32.5 million code written: ~7696 lines marginal compute: ~$114 +- $30 human cost: $4500 (288 participants) "Closing the Empirical Loop: Autonomous AI Agents Conduct End-to-end Research With Human Participants" @advportfolio
12
15
6
99
188,191
Rishi Taparia retweeted
Huge congrats to @GarudaVC portfolio company C1! The agentic identity platform for people and agents. Identity is a critical problem in an agent-first world, and C1 built the leading platform to help enterprises govern agents securely while enabling businesses to move fast with agents. Incredible work by founders @alexbovee and @pquerna and the whole team. Thank you for letting us partner with you on the journey.
We’ve signed an agreement to acquire C1.ai, an agentic identity platform for people and AI agents. With C1.ai, Workday will become the identity governance layer for people, money, and agents. w.day/3UsJttH
1
1
13
798
Rishi Taparia retweeted
Silicon Valley keeps building me a husband when what I really want is a wife lol I don't want an agent that only does one-off, random tasks that are nice but not life-changing: booking flights, making dinner reservations, or buying tickets I want an AI wife who knows what’s on the family calendar, remembers there’s a birthday party Saturday and we haven’t bought a gift then sends me options, realizes we’re almost out of diapers and buys them, knows which bills are coming up, remembers someone needs a dentist appointment and books it, figures out what the hell we’re eating all week and what we need from the grocery store, and generally keeps track of the 47 things that need to happen before anyone else even realizes they need to happen. So, who's building me an AI wife that carries the mental load instead of an AI husband that waits for me to give it a chore??
659
428
265
6,693
869,870
Rishi Taparia retweeted
If AI agents spends more time waiting than generating answers, what should we actually be accelerating? And what does that mean for the next generation of AI chips? A look at where the time, memory and money go when agents get to work.
7
10
8
75
26,001
Rishi Taparia retweeted
This Substack essay deserves to be more viral than the video that spawned it. What a remarkable mom and dad, and what a remarkable portrait of a loving marriage. Everyone should read this; it is truly encouraging.
The comments section and TikTok commentariat did their thing shaming a dad at a Giants for eating while his wife dealt with one of their kids. His wife has now posted a rebuttal: themomoftheyear.substack.com…
13
72
2
548
96,494
Rishi Taparia retweeted
Whenever someone tells me they're expecting their first baby and they're nervous, I tell them the following: 'Oh my goodness, that's wonderful. I am so happy for you. Listen, of course you're nervous but here's the deal: you're ready for all the bad stuff. You've been very tired before. You've been in pain before. You've been worried about money before. You've felt like an incapable moron before. So you'll be fine with the difficult parts! You're already a pro. What you're NOT ready for is the wonderful parts. NOTHING can prepare you for how amazing this will be. There is no practice for that. (via Rob Delaney, A Heart That Works)
you CANNOT understand how cute your kids will be to you. it is incomprehensible. I spend 50% of my free time looking at photos and videos of my children and just feeling really really really good. they are SO AMAZING. and so cute! 🤩
37
765
26
15,780
1,239,608
Rishi Taparia retweeted
Today we are releasing our speculative decoding implementation in our inference engine uzu. Initially for Qwen3.6 27B, with support for Qwen3.8 27B and Muse Glimmer coming soon. On Apple M5-series chips, we outperform MTPLX (MLX + speculative decoding) by almost 2x, and llama.cpp by over 3x at comparable quantization levels, with the strongest gains achieved on mathematical reasoning and coding tasks. Run the model: Mirai-M: trymirai.com/local-models/al… Mirai-L: trymirai.com/local-models/al… Explore the benchmarks: trymirai.com/metrics Learn more about our speculative decoding implementation: trymirai.com/blog/speculativ… Our draft model, quantized checkpoint format, verification algorithm, and GPU kernels are co-designed from the ground up around the latest Apple M5 chips to take maximum advantage of GPU Neural Accelerators. Unlike popular speculative decoding architectures such as model-native MTP, which produce small draft chains of 3-4 tokens at a time, we use extremely aggressive speculative budgets of 16-32 tokens. This enables us to use Neural Accelerator-backed GEMM kernels, achieving maximum utilization of hardware arithmetic throughput.
30
61
20
463
223,188
Rishi Taparia retweeted
Today, we’re excited to unveil Heaviside-1, @arenaphysica's second generation foundation model for electromagnetism. 5 months since the release of Heaviside-0, Heaviside-1 is a crucial milestone on our quest to build EM superintelligence: a foundation model that understands EM across the spectrum, from RF through photonics, capable of designing the next generation of electronics. Major updates: - Heaviside-1 is >10x the size of Heaviside-0 (roughly the size of GPT-2), trained on 250k unique designs with >500B unique EM field samples. - It runs 10^5 x faster than commercial solvers, with accuracy <1 dB. - Heaviside-1 natively encodes 3D structures, their material properties, and excitation patterns as input. It predicts full EM fields at arbitrary locations in space, not just downstream quantities like S-parameters. - Most importantly, our bet on fields lets Heaviside generalize OOD proving that it’s learning real EM physics: Heaviside-1 generalizes outside of distribution very well, leaping from 0.99 dB to 0.53 dB S-parameter error. Try it out in Atlas Fields Studio in beta today (link in replies). I personally had a lot of fun seeing how EM fields twist and curl around different circuits. I wish I had this when I was learning EM in college. I hope you enjoy it as much as we did.
18
28
17
195
60,053
Rishi Taparia retweeted
We’re at the very beginning of a huge overhaul of the EDA space. Cadence and Synopsys are gonna have to change their business model for sure, but there’s also a lot of AI-heavy startup teams trying to optimize single slices of the chip design process (e.g. RTL gen or verification). My prediction is Cadence and Synopsys will buy a few of these to try to keep up, but there’s room for a few companies who focus on the end to end design flow to fully disrupt the industry
8
18
5
257
74,199
Rishi Taparia retweeted
if this stuff interests you, poke around the project site with all the details: yoheinakajima.github.io/synt… also shoutout to @ExScienceAI for helping me polish my papers :)
2
1
3
2,874
Bernard Arnault going direct is remarkable and worth reading in full The single most impressive thing is that Arnault (personally, not via company spokesman) responds to a major hit piece without at all accepting the role of defendant He never complains about “biased reporters” or “they ignored our fact check” etc. He instead: 1) Does not flinch. He starts by giving the rebuttal a title: “Thank You.” Not “Addressing Recent Allegations” or whatever but implying the expose has backfired and made him look good. Eg: “They used to call me The Terminator. I prefer ‘the last royal family.’ It’s more elegant, and probably better for Dior sales.” Tone stays measured and bemused throughout; it’s giving Logan Roy 2) Lampoons the allegations. His wry humor makes the expose seem silly and theatrical, his tone is like he’s divulging a stupid child. He opens with, “Apparently I am the head of ‘the last royal family of France’.” And drops gems like, “At the Arnault family, apparently, we don’t argue, we conspire. We don’t deliberate, we plot. It’s novelistic… What, in an ordinary family, is simply called Sunday lunch becomes, in our case, an intrigue. It’s a matter of vocabulary.” 3) Reframes the whole narrative. From billionaire to steward, from an oligarch pursuing power to a patriarch preserving an institution built by and for others Eg: “You describe power as though it were a prize to be won. I experience it as a responsibility to be passed on. The houses entrusted to me are older than I am. Most will outlive me. My role is to hand them on stronger than I found them. That is why I reject the idea of a dynasty. Dynasties exist to preserve privileges. We work to preserve craftsmanship.” Interestingly, in the very act of doing this Arnault shows how masterful he is at wielding power, and he summons that skill here to make the reporters look like bumbling neophytes
Merci : la réaction de Bernard Arnault à « l’enquête » du Monde
71
328
56
3,363
652,030
Rishi Taparia retweeted
Replying to @taps
@taps and I @GarudaVC hosted our annual meeting in SF a few weeks ago for our LPs and invited guests, followed by our community-wide Spring Celebration right after. It's always a fun opportunity to share where we're investing and what we're learning, having a few founders and leaders share insights on where their markets are going, and bringing our growing community together for some fun! Special thanks to portfolio founders @adarshkulkarni from @FoundryRobotics, @shaneehrhardt from @ExScienceAI , and Derek Ben and Michael at Vardera - who were busy closing deals the morning of our event! - for joining us. And thank you to the inimitable @rodriscoll from @scalevp and Jonathan Roosevelt from Goldman Sachs for sharing their perspectives and wisdom. Thanks to our amazing sponsors Burkland, Neutech, Inc., Fenwick, and Citizens!
1
5
240
Rishi Taparia retweeted
Mirai Tech is now Mirai Labs (@trymirai) A full-stack frontier lab for on-device AI. Models, inference engine, quantization, and application layer, all co-designed from the hardware constraint up. Interactive AI is a continuous loop: parsing, validating, tool calls, rendering. That only works when latency disappears. Cloud models can't get you there as they're built for large batches, abundant memory, throughput as the metric. On-device inverts every assumption, requiring the architecture that was built from the hardware constraint up. The silicon is shipped. The software stack to match the chip doesn't exist yet. That's ours to build.
3
7
1
49
164,788
Rishi Taparia retweeted
“@arenaphysica’s bet is that the same idea works as a foundation model rather than a one-off demo: faster inference, trained on roughly 10 million geometries plus decades of simulation data, with lab measurements added as a third data leg to help close the sim-to-reality gap.” Thanks to the @hardwarefyi team for highlighting @m__frei’s CDFAM talk in this week’s newsletter. Read the full piece here: hardwarefyi.substack.com/p/r….
5
16
1,670