• Two years ago (March 2023): Published a blog post titled "The End of Front-End Development" after OpenAI released GPT-4. Skeptical of claims that human developers would be made redundant, believing LLMs would augment rather than replace them.
• Companies and AI usage: Companies have been adopting AI tools more. Forbes reported AI writes over 25% of code at Google, but it's not acting independently. A skilled human is in control, guiding and editing the AI's output. Startups claiming their AIs can fully replace human developers have faced problems. Every AI success story still has skilled human developers as a necessary part.
• Drifting off-road: Experimented with AI tooling like Cursor's "agent" mode with Claude Sonnet. It's remarkable for some tasks but not perfect and requires guidance. Like driving with "cruise control", it can drift if not corrected. No-coders also face limitations with LLMs. There are tasks LLMs aren't good at, and one needs to know when to delegate and when to code manually.
• The current job market: Tough job market with fewer high-quality job listings and heavy competition. Not because companies are replacing developers with AI but due to macro-economic factors (high interest rates), layoffs in big tech companies, and AI myths (belief that AI will make developers obsolete).
• Looking forward: Two years later, companies still need human developers, and AI evangelists still claim otherwise. AI models will continue to improve but progress is becoming more incremental. Tech jobs have been increasing over the past year, suggesting the market is trending in the right direction.
• Concerns: Confident that AI isn't taking software jobs yet. Worried about the next generation of developers getting lulled into a trance when using LLM agents and not building debugging skills. But using LLMs proactively can be helpful. Nobody knows how the next few years will play out, but a "developer renaissance" might happen when companies realize the need for human developers.