Reading Data From DNA: Why It's Still Unsolved in 2026

Storing data in DNA works. Reading data from DNA back exactly, cheaply and fast is still unsolved. Here's how DNA data storage works, why trace reconstruction from noisy reads is so hard, and the fair scoreboard nobody has built yet. THE OPEN PROBLEM: Which method rebuilds DNA strands best — when everyone is tested fairly? CHAPTERS 0:00 Hook 0:42 DNA as a hard drive 1:32 The problem: messy copies 2:47 Why this matters 3:54 How far we are today 4:42 What people have tried 6:37 The hidden problem: no fair scoreboard 7:26 What we don't know yet 8:04 Who's working on it 8:51 The sources anyone can use 9:27 How you could help A NOTE ON THE NUMBERS • The per-read failure rates for TReconLM (~65% at 10 reads on Noisy-DNA) and alignment-and-vote (58% at 5 reads on Microsoft data) are from the TReconLM paper's appendix tables, as cited in the Helix Vault case study. • The belief-combining method (arXiv 2601.18920) is a preprint. • Codec bar lengths in the "Why this matters" chart are illustrative; the measured figures are stated on screen. SOURCES • Weindel, Girsch & Heckel, "Trace Reconstruction with Language Models" (TReconLM), TU Munich — https://arxiv.org/abs/2507.12927 • Gimpel et al., "Comparison of state-of-the-art error-correction coding for sequence-based DNA data storage" (ETH Zürich; Nature Communications 2026) — https://www.biorxiv.org/content/10.1101/20... • DNAformer, Technion (Nature Machine Intelligence, 2025) — coverage: https://www.synbiobeta.com/read/ai-superch... • Gu et al., "Efficient trace reconstruction in DNA storage systems using Bidirectional Beam Search" (A*STAR; iScience 2025) — https://www.biorxiv.org/content/10.1101/20... • Nouri, belief-combining reconstruction (preprint) — https://arxiv.org/abs/2601.18920 • Microsoft Research, Trellis BMA and the Clustered Nanopore Reads dataset — https://www.microsoft.com/en-us/research/?... • IARPA MIST program — https://www.iarpa.gov/research-programs/mist • Twist Bioscience spins out Atlas Data Storage ($155M seed, 2025) — https://investors.twistbioscience.com/news... • DT4DDS simulator and dt4dds-benchmark (ETH Zürich) — GitHub • DNA Data Storage Alliance (SNIA) — https://www.snia.org/dna The Open Problem — Exploring the Questions Humanity Hasn't Answered Yet.