I love LLMs, I hate hype
geohot.github.io
July 12, 2026
3 min read
🔥🔥🔥🔥🔥
62/100
Summary
The author expresses enthusiasm for advancements in AI, particularly in LLMs, self-driving cars, video generation models, and coding agents. A recent setup of a Linux box with opencode on GLM-5.2 successfully executed commands, indicating progress in the usability of Linux systems.
Key Takeaways
- The author expresses excitement about advancements in AI, including LLMs, self-driving cars, and coding agents.
- The author criticizes the negative hype surrounding AI, arguing it is designed to instill fear and promote a false narrative of impending doom.
- The author believes that the value created by AI will not be captured by frontier labs, as progress is primarily driven by Moore's law and general advancements in computing.
- Programming is evolving, and while AI models can enhance productivity, they also require new skills and can lead to cognitive fatigue.
Community Sentiment
Positives
- The frontier models are a no-brainer at current subscription prices, making advanced AI accessible to many companies that previously couldn't afford it.
- We're entering a 'have it your way' era, where forking open-source projects to fit specific needs is easier than ever, potentially spurring innovation.
- Healthy competition is emerging in the AI space, making it likely that local models will become more accessible and affordable for individual users.
Concerns
- Many commenters are skeptical about the sustainability of current token rates for frontier models, fearing a future price hike.
- There's a pervasive sense that frontier labs are more focused on capturing value than genuinely improving model performance, leading to concerns about their practices.
- The quality of some models is questioned, with users feeling that many are just 'good enough' rather than truly exceptional, reflecting a concern over diminishing returns in AI advancements.
