AI Coding Productivity: Cracks in the Bull Case

AI Coding Productivity Trial

• 💡 A real-world trial by AI nonprofit Meter found that seasoned engineers were 19% slower when using AI coding tools like Cursor.
• ⚠️ Instead of speeding up tasks, the AI often provided suggestions that required time-consuming corrections, undercutting the narrative of supercharged white-collar efficiency.

Nuances in AI Tool Benefits

• 🎯 Prior studies showed AI tools benefit junior engineers for simpler, well-scoped tasks, helping them level up.
• 🧠 However, the latest study suggests AI may increase reliance on senior talent for debugging, refining, and shipping the final product.
• 💰 This reliance on senior talent helps explain the talent wars and high offers for top AI engineers.

Broader AI Adoption Trends

• 📈 New data from RAMP shows paid AI tool usage is stalling, remaining flat at around 40% after a steep runup.
• 📉 In aggressive sectors like tech and finance, there was a slight pullback in AI tool usage.
• 📊 While adoption was explosive in early 2023 (1 in 20 companies), it has grown to nearly one in two companies on RAMP spending on AI tools.

AI's Role in Code Development

• 💻 Google and Meta report that around 50% of their code is now written by AI.
• 🧩 The payoff for AI tools may be more uneven and less immediate than initially hyped, potentially plateauing.
• 🚀 The push for AGI (Artificial General Intelligence) aims for models smarter than humans to handle a larger percentage of coding tasks.
• ⚠️ A potential danger is junior engineers accepting incorrect code, creating more work for senior engineers later if not checked early.