OpenAI Sol Nearly Decodes Hidden Ghost-Font Message After 40-Minute Visual Reasoning Test

On 13 July 2026, OpenAI Sol model caught attention after spending nearly 40 minutes trying to decode a hidden Ghost Font message during a visual reasoning test. The experiment showed how much AI vision has improved while revealing the limits that still exist.

Unlike ordinary text, Ghost Font hides words inside moving visual patterns that people can recognise more easily than AI systems. The result offers another way to measure progress in multimodal intelligence, AI security and visual reasoning.

What Happened in the 40-Minute Ghost Font Challenge?

The experiment explained

Ghost Font is a visual technique that hides text inside moving dot patterns instead of regular letters. The hidden message only appears when those dots move together over time. Looking at a single frame reveals nothing more than random noise. In early July 2026, developers used this system to test OpenAI Sol in a lengthy visual reasoning session. The model spent nearly 40 minutes analysing the animation and trying different ways to uncover the concealed text.

JUST IN: OpenAI resets high-compute mode prompts and rolls out fixes after a desktop revamp caused usage-quotas confusion and workflow hiccups for Codex and ChatGPT Work. Could impact developer tooling adoption and workflow reliability. $OPENAI pic.twitter.com/b6KcSEjpyE

— Bpay News (@bpaynews) July 11, 2026

Why did OpenAI Sol attract attention?

The experiment gained interest because Sol came very close to decoding the hidden message but never reached the correct answer. Instead, it identified a built-in decoy message. That outcome showed how much multimodal AI has improved while reminding researchers that some visual reasoning problems remain difficult.

How Ghost Font Works and Why AI Finds It Difficult?

How does Ghost Font reveal hidden text?

Ghost Font does not use traditional typography. It creates groups of moving dots that travel in different directions. The human brain naturally combines these movements and recognises the hidden letters. Most AI vision models process images one frame at a time. Since no individual frame contains readable text, finding the real message becomes much harder for the model.

Why do decoy messages confuse AI?

Ghost Font includes a fake hidden message alongside the real one. Many AI systems detect this decoy first and stop looking any further. That makes Ghost Font a useful way to evaluate visual reasoning rather than simple image recognition. Researchers also point out that Ghost Font is not a form of encryption.

A system built specifically to analyse video sequences could still recover the original message. Even so, the experiment exposes a weakness that current multimodal AI models still face.

What does this test say about OpenAI Sol’s Visual Reasoning?

Is OpenAI Sol better than earlier AI models?

The Ghost Font experiment suggests that OpenAI Sol can spend more time reasoning through difficult visual problems than many earlier AI models. OpenAI has also reported improvements in visual reasoning systems that can zoom, crop, and inspect images while solving more complex tasks. Its latest visual reasoning model achieved 95.7% accuracy on the visual search benchmark, showing stronger image understanding than previous approaches.

OpenAI says GPT-5.6 Sol comes within 1 point of Claude Fable 5 while running 61% faster at roughly half the estimated cost. The enterprise AI question is shifting from “Which model is smartest?” to “Which workflow burns the fewest tokens and still works?” pic.twitter.com/b0N6azJxj4

— David Eberle (@davidaeberle) July 12, 2026

Why does “nearly decoding” still matter?

Coming close without solving the puzzle still tells researchers something useful. The experiment suggests that AI can work through difficult visual problems instead of giving up after a quick attempt. Better visual reasoning could support applications in medical imaging, robotics, autonomous vehicles and scientific research, where understanding complex visual information matters more than recognising simple objects.

Could Ghost Font Become an AI-Resistant Communication Tool?

Ghost Font remains an experimental project, but researchers see several possible uses for it. It could help build stronger CAPTCHAs, reduce automated content scraping, and provide a tougher benchmark for testing AI vision systems. It may also support digital watermarking and copyright protection.

As AI models improve, experiments like Ghost Font give researchers another way to measure how well machines understand visual information rather than simply reading images.

Conclusion

The Ghost Font experiment shows that AI vision continues to improve, but human-like visual reasoning is still a difficult problem to solve. OpenAI Sol came close to decoding a message hidden through motion rather than static images, yet it missed the correct answer. Tests like this help researchers measure progress and identify the areas where AI still needs better visual reasoning.

Disclaimer:

The content shared by Meyka AI PTY LTD is for research and informational purposes only. Meyka is not a financial advisory service, and the information provided should not be treated as investment or trading advice.