Ancient Script Decoded By AI In Breakthrough Tech Feat
By 813 Staff
A closely watched product launch reveals Ancient Script Decoded By AI In Breakthrough Tech Feat, according to Boris Cherny (@bcherny) (on June 19, 2026).
Source: https://x.com/bcherny/status/2068064304503660962
The race to open-source history’s deepest secrets now runs through a command line. At stake is not just academic prestige, but credibility for the AI coding assistants that major tech firms are betting their developer ecosystems on. If a tool like Claude Code can crack a 3,500-year-old undeciphered script, the implications extend far beyond linguistics—it signals a new class of agentic software capable of genuine research. If it fails, the hype around AI replacing human expertise takes a significant hit.
Boris Cherny, an engineer well-known in the TypeScript and infrastructure communities, posted a provocative use case on X this week. Cherny, who previously helped build tools at Meta and has deep ties to San Francisco’s engineering scene, shared that he is using Claude Code (Anthropic’s agentic coding tool) to decipher Linear A, the writing system of the Minoan civilization that has resisted all conventional decryption attempts for over a century. The tweet, posted on June 19, 2026, shows a command-line session where the agent processes Linear A tablets and attempts to map symbols to known phonetic values. Internal documents from Anthropic indicate the company is closely watching these unconventional integrations, though they have not officially partnered with any archaeological institutions.
The rollout of Claude Code, compared to competitors like GitHub Copilot Workspace, has been anything but smooth. Engineers close to the project say the agent’s ability to chain research steps—from OCR scanning of tablet images to cross-referencing Linear B deciphered patterns—is impressive, but it hallucinates phonological mappings frequently. Cherny’s results are still unverified by any specialist in Mycenaean paleography, and the broader academic community remains skeptical of AI-driven linguistic analysis. What is clear is that the agent is processing the corpus faster than any human team could, generating candidate translations that would take classical scholars years to propose.
The next step is peer review. The output from Cherny’s session will need to be validated against known vocabulary from related scripts. If the approach holds, it could be turned on other undeciphered languages, including Proto-Elamite and Indus Valley script. For now, the tech industry is watching whether a CLI tool can do what the world’s top linguists have not.
