Jan Ole Ernst

Design verification still depends on engineers manually reading hundreds of pages of JEDEC specs and translating them into testable representations. We've released DRAMBench, an open benchmark for autoformalizing DRAM specs and the paper behind it, presented at the @iclr_conf VerifAI Workshop. Our approach introduces an intermediate formal layer of timed Petri nets that capture device states, commands, and timing constraints in a compact, executable model. From that single representation, verification collateral derives automatically. By Jan Ole Ernst, Dmitri Saberi, Derek Christ, Thomas Zimmermann, Rajath Salegame, Suhaas Bhat, Stanislav Levental, Thomas Ahle, and Matthias Jung, in partnership with @FraunhoferIESE. Both DRAMBench and DRAMPyML are open source, Apache 2.0. Read it here: arxiv.org/abs/2605.00058

Code is open source here: github.com/normal-computi…

I will be presenting this work tomorrow at @iclr_conf Verif-AI 2 Workshop. Drop by if you are interested in AI & chip verification/design.

At @iclr_conf in Rio? Join us for the Normal Silicon Social on Sunday, April 26 from 6:30 to 8:30 PM at the Copacabana Palace. Request an invite: luma.com/3ioi64wv #ICLR2026

Normal Computing has raised $50 million in strategic funding led by @SamsungCatalyst, bringing total funding to more than $85 million. We build AI for the semiconductor industry and are developing a new class of computing hardware, using our software to design our own hardware IP. We're partnered with more than half of the top ten semiconductor companies by revenue through Normal EDA, our purpose-built AI platform. In parallel, we completed the tape-out of CN101, the world's first thermodynamic computing chip, targeting up to 1000x gains in energy efficiency for AI workloads. Investors include @GalvanizeLLC , @CelestaCapital , @drivecapital , Eric Schmidt's First Spark Ventures, Micron Ventures, Brevan Howard Macro Venture Fund, and @ArcternVC. We're hiring: careers.normalcomputing.com Full announcement: prnewswire.com/news-releases/…

my results on AI for autonomous progress: we ran codex+claude for 43 days straight to build a System Verilog compiler/simulator normalcomputing.com/blog/building-…

The format has remained fairly consistent over the last decades.. curious how that will change in the future.

My PhD thesis just got published… was kinda fun looking back at how it shaped over time. ora.ox.ac.uk/objects/uuid:b…

Claude Code is the Inflection Point, What It Is, How We Use It, Industry Repercussions, Microsoft's Dilemma, Why Anthropic Is Winning. newsletter.semianalysis.com/p/claude-code-…

Today, the team at Normal Computing announces that we have achieved first silicon for CN101: the world’s first thermodynamic computing chip. CN101 accelerates AI inference, large-scale linear algebra, and stochastic sampling for diffusion models, marking a key step toward up to 1000× more energy-efficient AI. By harnessing the intrinsic dynamics of physical systems, CN101 maximizes compute output within fixed datacenter energy budgets. “Thermodynamic computing has the potential to define the next decades’ scaling laws by exploiting the physical realization of AI algorithms. Achieving first silicon success is a historic moment for this emerging paradigm.” –Faris Sbahi, CEO

me, trying to program on a flight with no wifi, forgot what its like to have no tab 🥵

📜 Full paper: arxiv.org/abs/2501.14372 This work was done in collaboration with @frtimlive and @HarmlessFlower7 who are both incredible.

🚀 Code is open-source & ready to run! Includes quantum environments, RL pipelines and additional utils for reproducing experiments. 🔗 github.com/jan-o-e/RL4qcW…

🚨 New RL4Physics Paper @icmlconf 🚨 🤖 Physics-constrained RL method that steers the evolution of quantum systems 🎯 Achieves SOTA realistic control of noisy, open quantum systems 🧠 RL discovers novel reset pulse for a critical quantum computing task Check it out ⤵️

🌌 Quantum control is essential for building quantum technologies. Traditional methods often struggle when systems are open, noisy, or too complex. Our approach embeds physical constraints directly into the RL process, enhancing both performance and computational efficiency.

@mo_baioumy 👀

I miss going to a real lab