# Eric Lu

Member of Technical Staff at Cognition: SWE-1.6/1.7 model training, FrontierCode benchmark lead author, RSA-260 factorization  
Cognition  
San Francisco, California, United States

## At Cognition

- Role: Member of Technical Staff (Engineering)
- Joined: Feb 2026
- How they joined: Directly
- Before Cognition: Member of Technical Staff, Imbue
- Overlapped before Cognition with other Cognition people at: Harvard University (https://www.autumn.ai/resources/cognition/worked-together)

## Identity
- LinkedIn: https://www.linkedin.com/in/penlu
- X: @penlume
- GitHub: https://github.com/penlu
- Website: https://penlu.me
- Domain: cognition.com

## Notes
- Factored RSA-260 (a 260-digit, 862-bit RSA Challenge number), announced 2026-09-03 on X with a 130-digit factor and the words "divides RSA-260"; the Cognition write-up describes Devin building a GPU lattice siever (modified CADO-NFS) run on spare cluster compute, ~4,900 GPU-days. (https://cognition.com/blog/factoring-rsa-260)
- Cognition blog, in his own words: "Over the past few weeks, the Cognition research team and I have been optimizing our job scheduler to better use disaggregated compute. As a proof of concept, and because I've enjoyed factoring numbers as a hobby for the past ten years or so, I drove a bevy of Devins to obtain a factorization of RSA-260.". (https://cognition.com/blog/factoring-rsa-260)
- Cognition blog: "my Devins built the world's highest-performance GPU lattice siever, which enables factoring numbers at 10x lower cost than the previous public state of the art." (https://cognition.com/blog/factoring-rsa-260)
- Listed first author of 'Introducing FrontierCode' (Jun 2026), Cognition's benchmark of mergeable, high-quality code built with 20+ open-source maintainers; the best model scored 13.4% on the Diamond subset. (https://cognition.com/blog/frontier-code)
- Joined Cognition as Member of Technical Staff (research) in Feb 2026, directly from Imbue rather than via Cognition's July 2025 acquisition of Windsurf. (https://www.linkedin.com/in/penlu)
- Factoring is a long-running hobby: he ran P-1 factoring for the Great Internet Mersenne Prime Search and held 4th place on the GPU to 72 largest-factor leaderboard as of Dec 2024, and published penlu/factoring for cracking FactorDB numbers. (https://penlu.me)
- Said the factoring was squeezed in around other work: "this was the high priority workload that I was fitting the factoring around. more to come!" alongside a post that visualizations were made with SWE-2. (https://x.com/penlume/status/2098101252374909104)
- Co-author on 'SWE-1.7: Frontier Intelligence at a Fraction of the Cost' (Jul 2026), covering RL entropy preservation, multi-cluster fault-tolerant training and self-compaction; on X he wrote "scaling this has been giga fun. more to come." (https://cognition.com/blog/swe-1-7)
- Co-author on 'Introducing SWE 1.6: Improving Model UX' (Apr 2026), a model shipped in Windsurf post-trained to cut overthinking, sequential tool calls and looping, at up to 950 tok/s. (https://cognition.com/blog/swe-1-6)
- Co-author on 'An Early Preview of SWE-1.6 and Research Update' (Mar 2026), which reports a 6x training speedup and SWE-1.6 scoring 11% higher than SWE-1.5 on SWE-Bench Pro. (https://cognition.com/blog/swe-1-6-preview)
- Before Cognition, MTS at Imbue (Feb 2025-Jan 2026): product engineering for Sculptor, a parallel background agent product, and automated rubric optimization for LLM bug-finding and instruction adherence. (https://www.linkedin.com/in/penlu)
- Posted that Devin rebuilt in Lean in under 15 minutes a formalization that took him much of a semester nine years earlier: "cost of formalization has fallen off a cliff."; links to his undergraduate Coq formalization of special relativity cited by HepLean. (https://x.com/penlume/status/2115806616796921885)
- Most recent Cognition-related post: "we continue to try to capture the true distribution of usage. there is still a long way to go. more to come!" (https://x.com/penlume/status/2115133981238898935)
- Credited the RSA-260 effort to tooling: "this effort was a testament not only to the prowesses and conveniences of Devin Cloud but also to the sophistication of CADO-NFS and the absurd strength of modern GPUs!" (https://x.com/penlume/status/2097787510219034812)
- Critiques LLM verbosity in public: "I am not aware of SWE benchmarks other than FrontierCode wherein writing more code will eventually reduce your score". (https://x.com/penlume/status/2080785298540396765)
- Harvard PhD in CS (2017-2024) advised by Eddie Kohler; dissertation 'Efficient Symbolic Execution and Reasoning for Low Level Code' (2024). (https://dash.harvard.edu/bitstreams/dc80f495-5245-4d5e-aef8-b16c026d2f0e/download)
- Worked in the Harvard OS-synthesis group with Jingmei Hu, David A. Holland, Ming Kawaguchi, Stephen Chong and Margo Seltzer: 'Trials and Tribulations in Synthesizing Operating Systems' (PLOS 2019), Aquarium/Cassiopea-Alewife tech report (2019), and 'Towards Porting Operating Systems with Program Synthesis' (TOPLAS 2023). (https://penlu.me)
- Public headline says "human computer interface". (https://www.linkedin.com/in/penlu)
- Widely covered press: John D. Cook, Wikipedia (Devin AI, RSA Factoring Challenge), and LinkedIn posts by Alex Pruden reported the RSA-260 factorization, which made him one of Cognition's most visible researchers outside the model releases. (https://www.johndcook.com/blog/2026/09/03/new-rsa-number-factored/)
- Co-authored 'Are Graph Neural Networks Optimal Approximation Algorithms?' (NeurIPS 2024) with Morris Yau, Nikolaos Karalias, Jessica Xu and Stefanie Jegelka (MIT); code at github.com/penlu/bespoke-gnn4do. (https://penlu.me)
- Presented 'Deferring Branches to Speed Up Symbolic Execution' with Eddie Kohler at the KLEE 2024 workshop at ICSE 2024. (https://conf.researchr.org/details/icse-2024/klee-2024-papers/30/Deferring-Branches-to-Speed-Up-Symbolic-Execution)
- Human-computer interface. I can operate your computers and make them do things fast.

In my PhD I applied symbolic execution and SMT solving for synthesis and verification of systems code. My BA is in computer science and physics.

Some other things I have previously worked on include: OS, machine learning, GPU programming, distributed systems, compilers; this comes from the profile summary and preserves the person's own positioning. (https://www.linkedin.com/in/penlu)
- Open-source work on GitHub (penlu): bespoke-gnn4do (NeurIPS 2024 code), factoring (FactorDB cracking tool), fastchem (chemprop GNN reimplemented in C/CUDA for better GPU occupancy). (https://penlu.me)
- X account @penlume (verified, ~12k followers) is his main public channel; bio: "human computer interface. naturally occurring feature of your environment". (https://x.com/penlume)
- BA in Computer Science and Physics, Harvard (2013-2017); internships at Reservoir Labs (2015, tensor decomposition on Broad Institute expression data) and Novus Partners (2014, Scala timeseries infra, 100-1000x cone-plot speedup). (https://www.linkedin.com/in/penlu)
- Previous roles include Member of Technical Staff at Imbue; Graduate Research Assistant at Harvard University; Engineer Intern at Reservoir Labs; Software Development Intern at Novus Partners, Inc. (https://www.linkedin.com/in/penlu)
- Education history includes Doctor of Philosophy - PhD, Computer Science, Harvard University; Bachelor of Arts - BA, Computer Science, Harvard University. (https://www.linkedin.com/in/penlu)
- His personal site links a second LinkedIn profile (linkedin.com/in/eric-lu-45868b2bb) in addition to /in/penlu. (https://penlu.me)
- Location is San Francisco, California, United States. (https://www.linkedin.com/in/penlu)
- Hobbies listed on his site: aircraft, Vocaloid trance, anime, and a computer-vision reconstruction and 3D print of a Hatsune Miku figurine. (https://penlu.me)

## Experience
- Member of Technical Staff at Cognition - San Francisco, California, United States (Feb 2026 - Present): research
- Member of Technical Staff at Imbue - San Francisco, California, United States (Feb 2025 - Jan 2026): - Product engineering for Sculptor, a parallel background agent product
- Automated rubric optimization for LLM-based bug-finding and instruction adherence
- Graduate Research Assistant at Harvard University - Cambridge, Massachusetts, United States (Sep 2017 - Dec 2024): PhD student under the supervision of Prof. Eddie Kohler.
- Engineer Intern at Reservoir Labs - New York, New York, United States (Jun 2015 - Aug 2015): Conducted applied tensor decomposition research:
- Applied tensor decompositions towards replicating results from Alzheimer’s gene expression study
- Clustered drugs based on Broad Institute RNA expression microarray data
- Extended C library with alternating Poisson regression tensor decomposition algorithm
- Contributed technical documentation and project result summaries
- Software Development Intern at Novus Partners, Inc. - New York, New York, United States (Jun 2014 - Aug 2014): Developed core services and infrastructure in Scala:
- Added support for future timeseries correlation matrices
- Improved cone plot asymptotic performance (producing 100-1000x speedup)
- Added specs2 test suites for key data structure
- Debugged MemSQL database queries during transition from MongoDB

## Education
- Harvard University - Doctor of Philosophy - PhD, Computer Science - (Sep 2017 - Dec 2024): Advisor: Eddie Kohler
Dissertation: "Efficient Symbolic Execution and Reasoning for Low-Level Code"
- Harvard University - Bachelor of Arts - BA, Computer Science - (Jan 2013 - Jan 2017): joint concentration in computer science and physics

Thesis: "A Formalization of Elements of Special Relativity in Coq"

## Sources
- https://www.linkedin.com/in/penlu
- https://cognition.com/blog/frontier-code
- https://cognition.com/blog/factoring-rsa-260
- https://cognition.com/blog/swe-1-7
- https://penlu.me
- https://en.wikipedia.org/wiki/Devin_AI
- https://www.highcircl.com/en/blog/devin-rsa-260-gpu-siever
- https://medium.com/@sohail_saifi/an-ai-coding-agent-just-helped-break-the-largest-rsa-number-on-record-c8ceb0cce77f
- https://blog.senthorus.ch/posts/rsa_260_factorization/
- https://www.linkedin.com/posts/alex-pruden_this-week-eric-lu-an-engineer-at-cognition-activity-7501715006670065664-fOMB
- https://en.wikipedia.org/wiki/RSA_Factoring_Challenge
- https://www.johndcook.com/blog/2026/09/03/new-rsa-number-factored/
- https://lilting.ch/en/articles/rsa-260-factored-how-computed
- https://x.com/penlume
- https://dash.harvard.edu/bitstreams/dc80f495-5245-4d5e-aef8-b16c026d2f0e/download
- https://conf.researchr.org/details/icse-2024/klee-2024-papers/30/Deferring-Branches-to-Speed-Up-Symbolic-Execution
- https://cognition.com/blog/frontier-code-1.1
- https://cognition.com/blog/swe-1-6
- https://cognition.com/blog/swe-1-6-preview
- https://github.com/penlu
- https://scholar.google.com/citations?user=MUvYvyEAAAAJ&hl=en
- https://x.com/penlume/status/2098101252374909104
- https://x.com/penlume/status/2115806616796921885
- https://x.com/penlume/status/2115133981238898935
- https://x.com/penlume/status/2097787510219034812
- https://x.com/penlume/status/2080785298540396765
- https://x.com/neal_wu/with_replies


Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-11-04. Profile: https://www.autumn.ai/resources/cognition/people/eric-lu · Everyone at Cognition: https://www.autumn.ai/resources/cognition/people