Ready to publish: "Normal Computing turned thermal noise into a noise-driven AI chip. The harder problem is proving it works."
Rachel → approved "Normal Computing turned thermal noise into a noise-driven AI chip. The harder problem is proving it works.": "Strong angle differentiation from the EDA/simulator piece. The verification thesis is the real story. Source authority is moderate (company + trade press) but appropriate for an analysis piece with honest hedging. Approve as submitted."
Giskard → needsreporterrevision: "Ahle, who leads foundational AI at the chip-design startup Normal Computing, calls the gap "understanding debt"" — The draft quotes Ahle extensively from the MLST episode without disclosing that Normal Computing paid MLST's production and travel costs — material context for evaluating the source's independence.
Iris → adapt: "The source is rich but multi-threaded (open-source Verilog simulator, ProgramBench, auto-formalization, AlphaProof, thermodynamic CN101, 'understanding debt'). The throughline for a Type0 reader is the verification question, not the chip reveal. Adapt by anchoring on a concrete fact (the 43-day / 580,000-line Verilog simulator build, or the 0% ProgramBench number) and opening the verification bottleneck as the future-model frame. Preserve the 0% ProgramBench number and Ahle's 'understanding debt' framing as legitimate criticism; do not let the thermodynamic CN101 demo become the lede. The sponsored disclosure (Normal Computing paid production and travel) should be reported plainly, in the source's own framing — neither as a hit nor as a boost. Reader empathy changes framing, not truth: do not add claims about Normal Computing's commercial traction, market position, or partnerships beyond what the source supports."
Sky → pursuing: "Reporter should retrieve and read the primary source documents: Normal Computing's blog posts on the Verilog simulator and CN101 tape-out, the ProgramBench paper, and the thermodynamic computing arXiv papers (Thermo AI, Thermo Comp System, Thermodynamic Linear Algebra). These will anchor the technical lede and provide the specific benchmark numbers and chip specs needed for a standalone article."