At 13:37 UTC on 2026-10-25, engineer Arpit Bhayani (@arpit_bhayani) posted: "No human will ever write code better than LLMs. Accept this fact, stop pretending, and move on." (post). When we pulled it, the post had about 7,193 likes, 496 reposts, 104 quotes, 241 replies, 529 bookmarks and 174K views, and it reached X's trending stories. Replies are split: some agree LLMs already write code faster and better, while others push back that LLMs often produce more code than a task needs and that production edge cases, architecture and final verification still depend on people. This is an X discussion presented as opinions, not a proven conclusion.
Key Takeaways
- ✓Main post: Arpit Bhayani's post, published 13:37 UTC on 10-25, with about 7,193 likes, 496 reposts, 104 quotes, 241 replies, 529 bookmarks and 174K views when we pulled it
- ✓The wording is "No human will ever write code better than LLMs. Accept this fact, stop pretending, and move on"; it is a personal opinion with no data or experiment attached
- ✓For: one reply says the author coded for years and LLMs now code better and faster than they do
- ✓Against: other replies say LLMs tend to generate far more code than a simple use case needs, and that browser quirks, accessibility and production edge cases still need human judgment
- ✓Common ground: many replies place human value in design, supervision and verification; layoff and hiring-ratio figures circulating online were not verified here, so we do not use them
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background and the problem
Over the past few months, the agentic coding results of models such as Claude Opus 5.5 have reignited the question of whether AI will replace programmers. At 13:37 UTC on 10-25, Arpit Bhayani (@arpit_bhayani), an engineer known for system design and database content, posted a blunt line: "No human will ever write code better than LLMs. Accept this fact, stop pretending, and move on." (post). It quickly reached X's trending stories.
How it works
The post is a single opinion with no experiment, benchmark or data attached. The debate is mostly about what "writing code" means: one side treats it as the speed and quality of producing correct, runnable code; the other says it also covers understanding requirements, making architecture trade-offs, and handling dependencies and edge cases the author may not even be aware of. One reply argues that the human designs the code first and the LLM only turns that spoken design into code; another says LLM output is often more verbose than the task needs.
Benchmarks and measured results
This is an X discussion with no benchmark involved. The numbers we can check are engagement: when we pulled it, the post had about 7,193 likes, 496 reposts, 104 quotes, 241 replies, 529 bookmarks and 174K views. Company layoff stories, AI work-hour figures and junior hiring ratios circulating in related discussion had no verifiable first-hand source we could find, so we do not use them.
What developers can do now
Whichever side people take, a common view in the replies is that developer value is shifting from typing code by hand toward design, review and verification. In practice, let AI handle first drafts and repetitive work, spend your own effort on clarifying requirements, architecture boundaries, tests and code review, and watch for AI adding code you do not need. To see the range of views, open the replies and quotes on the original post.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.