John Guibas

In October and November 2017 he and Tejpal Virdi published nine posts on machine learning and computer vision on twopointseven, a blog that the two ran together. The posts are dated from October 5 to November 25, 2017, and cover facial recognition through eigenfaces, compositional pattern-producing networks, convolutional neural networks, a simple neural network, generative adversarial networks, neural style transfer, computer vision for an autonomous car, a super-resolution generative adversarial network for medical images, and synthetic medical images from dual generative adversarial networks. The blog names the two as its authors and gives a GitHub link for each.⁠[1]

The post dated November 25, 2017 is titled “Synthetic Medical Images from Dual Generative Adversarial Networks” and is signed by John Guibas and Tejpal Virdi. A paper of the same title was submitted to arXiv on September 6, 2017 under the author names John T. Guibas, Tejpal S. Virdi and Peter S. Li, and its third version is dated January 8, 2018; it proposes generating synthetic medical images with two generative adversarial networks used in sequence, tested on retinal fundus images, and its authors started SynthMed, an online repository of synthetic medical images. His earlier personal homepage, whose news items run to March 2022, lists the paper among his publications, with the authors J Guibas, T Virdi and P Li and the venue given as Conference on Neural Information Processing Systems (NIPS ML4H), 2017.⁠[2][3][4]

His earlier personal homepage describes him as a junior at Stanford University studying computer science and statistics who did research on database systems and machine learning at the Stanford DAWN Lab, working with Matei Zaharia and Tatsu Hashimoto. It adds that he was then deep into crypto and interested in topics such as zero-knowledge systems, decentralized finance and blockchain infrastructure.⁠[4]

Three items in that homepage’s publication list are database papers that he co-wrote with Daniel Kang and others: “Accelerating Approximate Aggregation Queries with Expensive Predicates” (D Kang, J Guibas, P Bailis, T Hashimoto, Y Sun, M Zaharia; VLDB 2021), “Task-agnostic Indexes for Deep Learning-based Queries over Unstructured Data” (D Kang, J Guibas, P Bailis, T Hashimoto, M Zaharia; SIGMOD 2022), and “Data Management for ML-based Analytics and Beyond” (D Kang, J Guibas, P Bailis, T Hashimoto, Y Sun, M Zaharia), which was then marked as in submission. The publication list on his later personal website, jtguibas.com, gives the last of these as 2023, Journal of Data Science, and gives the title of the SIGMOD 2022 paper as “TASTI: Semantic Indexes for Machine Learning-based Queries over Unstructured Data”.⁠[4][5]

On November 24, 2021, the paper “Adaptive Fourier Neural Operators: Efficient Token Mixers for Transformers”, of which he is first author, was submitted to arXiv; his co-authors are Morteza Mardani, Zongyi Li, Andrew Tao, Anima Anandkumar and Bryan Catanzaro, and the second version is dated March 27, 2022. The paper proposes the Adaptive Fourier Neural Operator (AFNO), which mixes tokens in the Fourier domain and frames token mixing as a continuous global convolution that does not depend on the input resolution; the abstract states that it outperforms self-attention on few-shot segmentation and handles a sequence size of 65k on Cityscapes segmentation. His earlier personal homepage lists the paper under ICLR 2022 and, in a news item of February 2022, records that this work on using Fourier transforms in transformers had been accepted to ICLR.⁠[6][4]

The Computing Research Association (CRA) published the list for its 2022 Outstanding Undergraduate Researcher Award on January 4, 2022; the list is divided into Awardees, Runners-Up, Finalists and Honorable Mentions, and he appears among the Finalists as a student of Stanford University. The news items on his earlier personal homepage record that he received the Barry M. Goldwater Scholarship in March 2022; the Nominees and National Scholars page of Stanford University’s undergraduate research website lists him as a Goldwater Scholar in Computer Science under the Goldwater Scholarship for Fellowship Year 2021-2022.⁠[7][4][8]

He writes on his personal website, jtguibas.com, that before Succinct he trained foundation models at NVIDIANVIDIANVIDIA is an American semiconductor and computing technology company that provides graphics processing units, artificial intelligence infrastructure, networking interconnects, and computing software, and is listed on Nasdaq under the ticker NVDA.Open the full entry and published research at NeurIPS and ICLR. The repositories pinned on his GitHub profile include NVlabs/AFNO-transformer (the official PyTorch implementation of the Adaptive FNO transformer) as well as stanford-futuredata/tasti and stanford-futuredata/abae.⁠[9][10]

He writes on his personal website, jtguibas.com, that he dropped out of school in 2022 for the Thiel Fellowship.⁠[9]

On November 23, 2022, the Zero Knowledge Podcast released episode 255, in which the hosts Anna and Tarun talk with Uma Roy and John Guibas of Succinct Labs. The show notes state that the two discuss their interest in zero-knowledge proofs, their work with 0xPARC and the goal of Succinct Labs, which was to provide proof of consensus through SNARK-based light clients; the conversation also covers the difficulty of building zero-knowledge light clients on EthereumEthereumEthereum is an open, general-purpose smart contract blockchain that launched in 2015, and its native asset is ether (ETH). The project was proposed by Vitalik Buterin and advanced jointly by eight co-founders; the network now uses proof of stake and centers on the EVM, the account model, a rollup scaling roadmap, and the open EIP process.Open the full entry, their first implementation, which linked Gnosis Chain to Ethereum, and how Succinct relates to the wider blockchain space.⁠[11]

Uma Roy wrote in a Succinct blog post of August 20, 2024: “my cofounder John and I met at the 0xPARC summer residency”; she called it one of the most generative times of her life, and the post does not give the year.⁠[12]

On February 8, 2023, the Thiel Foundation announced the 20 members of the new class of the Thiel Fellowship; he was seventh on the list, with Succinct as his project and Palo Alto, CaliforniaCaliforniaCalifornia is a U.S. state on the Pacific coast; its capital is Sacramento and its largest city is Los Angeles. California was admitted to the Union on September 9, 1850, as the 31st state. Its population at the 2020 Census (April 1, 2020) was 39,538,223, the largest of any state.Open the full entry as his hometown. The announcement describes him as building a trust-minimized way for any blockchain to access Ethereum. It states that the Thiel Fellowship, launched in 2011, gives each fellow $100,000 over two years and that fellows must pause their college enrollment and work full-time on their technology or company; the new class joined 251 previous fellows, and more than a third of that year’s applicants had never applied to college.⁠[13]

On March 21, 2024, Succinct announced that it had raised a combined $55 million in seed and Series A rounds led by ParadigmParadigmParadigm is a technology investment firm founded in 2018 by Matt Huang and Fred Ehrsam, investing in crypto assets, artificial intelligence, robotics, and other emerging technologies at stages extending from initial funding to public markets. The firm also conducts protocol research, open-source software development, and company incubation.Open the full entry. The announcement on Succinct’s blog, signed by Uma Roy, lists Robot Ventures, BanklessBanklessBankless is a crypto media outlet focused on Ethereum and decentralized finance, run by Ryan Sean Adams and David Hoffman; its main products include a newsletter of the same name, a podcast, YouTube shows, and paid membership content. It advocates “going bankless” with crypto tools and has given rise to the community organization Bankless DAO and the early-stage venture fund Bankless Ventures.Open the full entry Ventures, Geometry and ZK Validator as the other investors, and Sreeram Kannan of EigenLayer, Sandeep NailwalSandeep NailwalSandeep Nailwal is an India-born entrepreneur and co-founder of the blockchain network Polygon (formerly Matic Network), who has served as chief executive officer of the Polygon Foundation since 2025 and launched a cryptocurrency relief fund for India during the COVID-19 pandemic.Open the full entry and Daniel Lubarov of Polygon, and Elad GilElad GilElad Gil is an entrepreneur, investor and author of High Growth Handbook. He co-founded Mixer Labs and Color, worked at Google and Twitter, runs the multistage investment firm Gil & Co, and co-hosts the No Priors podcast with Sarah Guo.Open the full entry among the angel investors. The Block wrote on the same day that Succinct Labs is based in San Francisco and led by its co-founder and chief executive, Uma Roy. AxiosAxiosAxios is a US digital news media company founded in 2016 by former Politico executives Jim VandeHei, Mike Allen, and Roy Schwartz and launched in January 2017. It is known for short news items and industry newsletters in the “Smart Brevity” style, especially for its Washington political coverage, and publishes local news newsletters in cities across the United States through Axios Local. Since 2022 it has been owned by Cox Enterprises.Open the full entry wrote on the same day that the new raise was $43 million from Paradigm, following an earlier round of $12 million from Paradigm in which Robot Ventures and Standard Crypto took part, for a total of $55 million.⁠[14][15][16]

SP1 (Succinct Processor 1) is a zero-knowledge virtual machine (zkVM) that Succinct announced on February 14, 2024; it verifies the execution of programs written in Rust or other languages compiled through LLVM, is open source under the MIT and Apache 2.0 licenses, and is built on open-source dependencies including Plonky3.⁠[17]

On August 6, 2024, he published “SP1 Benchmarks: 8/6/24” under his byline on Succinct’s blog. The post states that SP1 v1.1.1, with its new GPU prover, has cloud costs up to 10 times lower than other zkVMs on blockchain workloads such as light clients and EVM rollups; it argues that RISC0’s latest benchmarks were misleading because they compared SP1’s CPU performance with RISC0’s GPU results, and says that Succinct’s own comparison used identical hardware. The post attributes the gap to a precompile-centric architecture, a two-phase prover for read-write memory, and proof-system efficiencies such as a lower blowup factor and LogUp lookups; it puts SP1’s proving cost at about a tenth of a cent per transaction for average Ethereum blocks and lists planned improvements, among them a new recursion system. It notes that the benchmarks can be reproduced from a public repository and acknowledges that the methodology gives only a simplified view of zkVM performance.⁠[18]

On January 27, 2025, he published the “SP1 Security Update” under his byline on Succinct’s blog. The post discloses two vulnerabilities in SP1 V3: a missing check in the recursive verifier that could let a partial execution verify as a complete one, which Aligned, LambdaClass and 3MI Labs disclosed to Succinct and which Succinct also identified independently while preparing SP1 Turbo, and a missing validation of chip ordering, which the Succinct team found while preparing SP1 Turbo. It adds that Lev Soukhanov and Onur Kilic had notified Succinct of a flaw in Plonky3, a core dependency of SP1. The post states that all three issues were fixed in SP1 Turbo, the latest production version, that the routers of the mainnet verifier contracts had been frozen, and that the Succinct Prover Network would stop supporting SP1 V2 and V3 from February 15, 2025; the post holds that direct use of the HALT syscall is not by itself a vulnerability and that the issues raised in submissions about RISC-V compliance have little to no impact; it states that Succinct had received an audit from Zellic of its RV32IM standards compliance and that all contributors had been compensated through bug bounty programs run by either Succinct or PolygonPolygonPolygon (formerly Matic Network) is a blockchain network and project founded in 2017, originally to scale Ethereum, and driven by organizations including Polygon Labs and the Polygon Foundation; its native token was upgraded from MATIC to POL, and in recent years it has shifted toward stablecoin payments infrastructure.Open the full entry.⁠[19]

On April 21, 2025, the paper “vApps: Verifiable Applications at Internet Scale” was submitted to arXiv; its authors, in order, are Isaac Zhang, Kshitij Kulkarni, Tan Li, Daniel Wong, Thomas Kim, John Guibas, Uma Roy, Bryan Pellegrino and Ryan Zarick, and its fifth version is dated April 30, 2025. The paper presents vApps, a development framework that simplifies building and deploying verifiable blockchain applications through a Rust-based domain-specific language and SDK, with modular tools that cover verification, proof generation and cross-chain connectivity and that automatically use specialized precompiles and hardware acceleration; the abstract states that benchmarks on zero-knowledge virtual machines show reduced cycle counts, faster proofs, higher GPU throughput and smaller proofs through recursion. His Succinct blog post of May 19, 2025 refers to it as Succinct’s recent paper and links to it, and his personal website, jtguibas.com, lists it first among his papers.⁠[20][21][5]

On May 19, 2025, he published “Introducing the Succinct Network Architecture and the $PROVE Token” under his byline on Succinct’s blog. The post states that Succinct is building a protocol on Ethereum that coordinates distributed provers to generate zero-knowledge proofs, forming a two-sided marketplace between provers and requesters; off-chain auctioneer components match requests to provers, on-chain contracts settle state roots and proofs, and the auctioneer does not hold users’ funds. The post describes PROVE as an ERC-20 token on Ethereum with a fixed supply of 1 billion, used for payments, staking, governance and prover incentives; provers must stake PROVE to take part in auctions, and staked tokens can be slashed when a prover misses a proof deadline. It adds that the initial auction is a reverse auction that assigns each request to the lowest bidder, and that the mechanism can transition over time to the proof contest mechanism set out in the whitepaper.⁠[21]

On May 20, 2025, Succinct introduced SP1 Hypercube, a next-generation zkVM that it said could generate proofs for Ethereum blocks in under 12 seconds. The Block wrote that Succinct’s internal tests covered 10,000 Ethereum mainnet blocks, of which 93% were proved in real time, with an average proving time of 10.3 seconds on a cluster of 200 Nvidia RTX 4090 GPUs; the verifier and Ethereum prover code had been open-sourced, and full deployment awaited an audit. The Block describes him in the article as Succinct’s co-founder and chief technology officer, and he said: “Real-time proving is the space race of zero knowledge, a technical and symbolic breakthrough that once felt out of reach.” He added that a year earlier this had been seen as science fiction and that it was now open source and generated Ethereum proofs in under 12 seconds, and said: “It’s our four-minute mile for ZK.”⁠[22]

On August 5, 2025, Succinct announced the mainnet deployment of the Succinct Prover Network and the activation of its native token, PROVE. The Block describes the network as a decentralized marketplace on Ethereum in which applications submit requests for zero-knowledge proofs and independent provers compete to fulfill them; Succinct put its use at more than 35 protocols, among them Polygon, Mantle, Celestia and Lido, with more than five million proofs generated from 1,700 programs and more than $4 billion in secured value. The Block gives his title as chief technology officer of Succinct Labs, and he said: “Our goal from the start was to make proving infrastructure accessible at internet scale,” and “Going from whitepaper to a working mainnet in eight months shows what’s possible when engineering and cryptography come together with focus.”⁠[23]

On October 8, 2025, he published “Upgrading the Succinct Prover Network: Private Proving Is Here” under his byline on Succinct’s blog. The post states that Succinct Private Proving generates proofs inside a trusted execution environment (TEE), so that neither the verifier nor the prover can see an application’s input data; it is built on SP1, which lets developers write private applications in Rust instead of custom circuits, and the uses the post names are private stablecoins, identity verification for credentials such as passports, and confidential computation on financial data.⁠[24]

The author page on Succinct’s blog lists 25 posts under his name, the earliest being “SP1 Benchmarks: 8/6/24” of August 6, 2024 and the most recent the post of October 8, 2025. Those in between include “Introducing OP Succinct Lite: ZK Fraud Proofs on the OP Stack” (February 20, 2025), “vApps: Verifiable Applications at Internet Scale” (April 24, 2025), “SP1 Is Bitcoin Ready: The Era of ZK on Bitcoin Has Begun” (May 8, 2025) and “Introducing SP1-CC: Unleashing the Full Power of the EVM” (July 17, 2025), together with posts on the adoption of Succinct’s technology by projects such as Mantle, ArbitrumArbitrumArbitrum is an Ethereum scaling technology and network family developed by Offchain Labs. Its Arbitrum One network uses optimistic rollups, Nova uses the AnyTrust data availability scheme, and the related technology is also used to build dedicated chains. Arbitrum combines off-chain execution with settlement on Ethereum and supports EVM and Stylus contracts.Open the full entry, World Chain, Katana, Galxe and Phala Network and announcements of provers joining the Succinct Prover Network.⁠[25][26]

As of November 2026, his personal website, jtguibas.com, states that he has spent the past few years building open-source cryptography at Succinct. Its Projects page lists two projects: SP1, a zero-knowledge virtual machine that proves the correct execution of programs compiled for the RISC-V architecture, and the Succinct Prover Network, a protocol on Ethereum that coordinates a distributed network of provers to generate zero-knowledge proofs.⁠[9][5]

As of November 2026, his GitHub profile shows the name John Guibas, the bio Succinct and the location San Francisco, California, and the first of its pinned repositories is succinctlabs/sp1.⁠[10]

He writes on his personal website, jtguibas.com, that he grew up in Palo Alto, California.⁠[9]

He writes on his personal website that a brief stint in academia left him with an Erdős number of 3.⁠[9]

He writes on the homepage of his personal website that he makes angel investments in companies such as Applied Compute, C5R, Stochastic Processes, Thalassa Robotics and Ellipsis Labs. The website’s Investments page lists Applied Compute, Ellipsis Labs, Sail Research, Vigil Labs, Seismic and Catalyst Labs.⁠[9][27]

Succinct: Official website: https://succinct.xyz/ (opens in a new window).⁠[14][9]

Thiel Fellowship: Official website: https://thielfellowship.org/ (opens in a new window).⁠[28]

Personal website: https://jtguibas.com/ (opens in a new window).⁠[9][10]

Earlier personal homepage (content as of March 2022): https://sushimax-xyz.pages.dev/ (opens in a new window).⁠[4]

GitHub: Public account: https://github.com/jtguibas (opens in a new window).⁠[10][9][4]

X: Public account: https://x.com/jtguibas (opens in a new window).⁠[9][10][11]