Breaking

Google Adds Skill to Analyze Ancient Inscriptions in Plain English on Antigravity

July 7, 2026 at 11:19 EDT

AI Agents
Foundation Models
Software Dev & Coding

Google DeepMind has released the "Predicting the Past Skill" on its Google Antigravity development platform, allowing researchers to study ancient Greek and Latin inscriptions in plain English. By grounding Gemini directly in DeepMind's specialist models Aeneas and Ithaca, the feature lets historians analyze inscriptions and search for large-scale patterns without writing any code.

July 7, 2026 · Google DeepMind

Talking to the Ancient World in Plain English

The new "Predicting the Past" Skill grounds Gemini in the Aeneas and Ithaca models on Google Antigravity — letting historians restore, date, and contextualize ancient Greek and Latin inscriptions through conversation, no code required.

73%

restoration accuracy for gaps of up to 10 characters

176K+

inscriptions in the training dataset

90%+

often missing from surviving fragments the AI helps fill

Training data at scale

Small for an LLM, vast for epigraphy

Fragment surviving

<10%

Text often missing

90%+

The bulk of an inscription is usually gone — AI restoration and parallel discovery aim to reconstruct what remains.

How the pieces fit

From specialist models to plain-English dialogue

Aeneas

Latin: restore, date, attribute (2025)

Ithaca

Greek: restore, date, locate (2022)

→

Predicting the Past Skill

Grounds Gemini on Antigravity — analysis by conversation

What it unlocks

Expert-level support and new insights via parallel discovery
Restores gaps of unknown length; accepts image input
Human + AI improves interpretive accuracy

Where caution remains

~170K training set is small versus general LLMs
Results depend on image and text quality
Deep interpretation still needs human verification

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.