Story perspectives

AI Breakthroughs Stalled: Reliability Issues Persist in 2025

1/24/2026

34 6

Summary

Story summary

In 2025, AI companies expected breakthroughs in generative AI agents, but progress stalled and reliability remains unattainable.
A recent paper argues large language models cannot perform complex tasks reliably.
Vishal Sikka, an AI expert, says AI aids simple tasks but isn't dependable for critical operations.
Harmonic, a startup, claims progress in AI coding reliability via mathematical verification methods.
OpenAI findings show hallucinations persist as even advanced models produce inaccurate information.

1 of 1

Story summary

In 2025, AI companies expected breakthroughs in generative AI agents, but progress stalled and reliability remains unattainable.
A recent paper argues large language models cannot perform complex tasks reliably.
Vishal Sikka, an AI expert, says AI aids simple tasks but isn't dependable for critical operations.
Harmonic, a startup, claims progress in AI coding reliability via mathematical verification methods.
OpenAI findings show hallucinations persist as even advanced models produce inaccurate information.