Redis Creator Says Stop Fighting AI: Programming Logic Is Being Reshaped

Salvatore Sanfilippo, the creator of Redis, wrote a blog titled “Don’t fall into the anti‑AI hype,” in which he openly declares that he has stopped resisting AI‑driven programming.

He explains that state‑of‑the‑art large models can now complete substantial sub‑tasks or medium‑scale projects, and that for most projects writing code manually is no longer the sensible option unless one does it purely for fun.

“For most projects, writing code yourself is no longer a wise choice, unless you’re doing it just for fun.”

To demonstrate this, Antirez lists four tasks he completed in the past week using AI (primarily Claude Code). Each task would have taken weeks by hand but only a few hours with AI assistance:

Refactor the low‑level library linenoise : AI modified the library to support UTF‑8 and created a test framework that simulates terminal character display, a job he previously deemed low‑ROI.

Fix the “ghost bug” in Redis tests : The bug involved transient TCP deadlocks and timing issues. Claude Code reproduced the problem, inspected process state, and fixed the bug.

Hand‑write a BERT C‑language inference library : He wanted a pure C runtime for BERT embeddings. Claude Code produced about 700 lines of code in five minutes, with performance only 15 % slower than PyTorch.

Rewrite the Redis Streams core : Feeding a design document to the AI, it recreated in twenty minutes the work he had spent weeks doing on the internal Streams implementation.

Antirez notes that after the AI reproduced his work, most of his time was spent reviewing and authorizing the generated commands.

He reflects on the shift in programming philosophy: previously developers needed deep knowledge of system architecture, syntax, memory management, and concurrency, while today success depends on the ability to clearly describe goals to an AI and to perform code reviews on AI‑generated output.

Old programming : Required mastery of low‑level details and extensive debugging.

New programming : Demands strong mental modeling to articulate tasks to AI and to oversee the results.

Antirez likens this transformation to a “democratization” similar to the 1990s open‑source movement, giving small teams or individual developers capabilities comparable to large companies.

He also expresses concern about AI compute concentration in the hands of a few giants and the potential for programmer unemployment, but his advice to developers is pragmatic:

“Regardless of your stance on AI, pure refusal and resistance are meaningless. Missing AI won’t save your career.”

Don’t validate bias with a five‑minute test : Avoid asking AI trivial questions just to mock its mistakes.

Invest weeks to deeply use AI : Integrate it into real projects to find a productive human‑AI rhythm.

Learn to “self‑replicate” : Leverage AI to amplify your output.

Comments on Hacker News echo his view, noting that even a master of hand‑written TCP stacks now relies on AI for C code, signaling the end of a “pure‑purist” era and the arrival of an “intent‑driven” programming age.