Neural Toons Original

AI's Brain Power: Not As Smart As You Think?

23 July 2026 · The Neural Feed
AI's Brain Power: Not As Smart As You Think? – Neural Toons comic

“Turns out, even AI needs to learn to connect the dots.”

The story behind the toon

What happened: A new benchmark, Relay-Bench, tests how well AI models can solve complex problems that require many steps across different topics like coding and math. The best AI model, GPT-5.5, only scored 43.3% on this test. This shows that current AI still struggles with long, multi-step reasoning and connecting information from various fields.

This Neural Toon is based on the story “Relay-Bench tests LLMs on multi-domain chains; GPT-5.5 scores 43.3%”.

Read our summary of the story · original source

Comic transcript

Panel 1 – Jordan: Yeah, it's like a git rebase gone wrong.

Panel 1 – Taylor: Jordan, my AI assistant keeps messing up long tasks!

Panel 1 – Fact: AI sometimes struggles with complex, multi-step tasks.

Panel 2 – Jordan: New 'Relay-Bench' test. It got 43.3%.

Panel 2 – Taylor: But GPT-5.5 is super smart, right?!

Panel 2 – Fact: GPT-5.5 (the best available AI) scored only 43.3% on a new tough test.

Panel 3 – Jordan: Not yet. Humans still needed for long chains.

Panel 3 – Taylor: So, AI can't do complex projects alone?

Panel 3 – Fact: Don't rely on AI for super complex, multi-step reasoning yet.

Panel 4 – Jordan: Like compiling code without dependencies.

Panel 4 – Taylor: It can't connect different topics?

Panel 4 – Fact: AI struggles with long, cross-domain reasoning chains.

More Neural Toons · Neural Snaps · The Neural Feed