Goalposts

Hacker News has set AI a lot of challenges over the years. Which ones has it met?

2025 Marchdeadbabe

About why some people remain unimpressed by LLMs despite their abilities.

Here’s when we can start getting excited about LLMs: when they start making new and valid scientific discoveries that can radically change our world.

When an AI can say “Here’s how you make better, smaller, more powerful batteries, follow these plans”, then we will have a reason to worship AI.

When AI can bring us wonders like room temperature semiconductors, fast interstellar travel, anti-gravity tech, solutions to world hunger and energy consumption, then it will have fulfilled the promise of what AI could do for humanity.

Until then, LLMs are just fancy search and natural language processors. Puppets with strings. It’s about as impressive as Google was when it first came out.

An AI makes new, valid scientific discoveries, such as giving working plans for better, smaller, more powerful batteries.

Has this happened?

Yes 37 (35%)Not sure 18 (17%)No 51 (48%)

Votes cast 1–2 October 2026: 100,590 votes from 9,694 people.