"Hold on, let me refine the prompt one more time…"
→ Take the AIBT TestBACKSPACE is the AI user archetype defined by extreme precision in input and patience in iteration. If you spend more time writing your prompt than the AI spends generating the response, if you've ever rewritten a prompt seven times before sending it, if you have a private collection of "good prompts" you've refined over months — you are almost certainly a BACKSPACE. The name comes from the keyboard key you use most: not because you make typos, but because you're constantly editing.
BACKSPACE users are the original prompt engineers. They were getting good results out of GPT-3 in 2021 when most people were getting nonsense, because they understood that the limiting factor wasn't the model — it was the prompt. They write prompts the way a programmer writes function signatures: precise types, explicit constraints, examples included. The output they get is correspondingly excellent.
BACKSPACE has the most distinctive prompt-quality fingerprint of any user type. Maximum precision, maximum patience, moderate-to-high on most other dimensions:
The 95% on Prompt is what defines BACKSPACE. They write detailed, structured prompts with role definitions, format specs, examples, and explicit constraints. They iterate patiently when the first response isn't right. They have nuanced opinions about which models work best for which tasks. The opposite of CTRL+V in almost every way.
This prompt is 80 words. A CTRL+V would have written "group dicts by key in python." Both prompts get a function. Only one gets the right function.
The output of round 3 is 10x better than the output of round 1. A less patient user would have given up after round 1 and used a worse tagline. BACKSPACE keeps iterating because they trust the process.
This is the BACKSPACE superpower: extracting genuinely good output by removing the AI's tendency to default to mediocre patterns. Every constraint they add eliminates a class of failure modes. The cumulative effect is professional-grade content.
BACKSPACE users get fundamentally different value from AI than other user types. While CTRL+V users use AI for speed, BACKSPACE users use it for quality they couldn't easily produce themselves. A BACKSPACE prompt is essentially a function call to a very flexible function — and the function returns exactly what was asked for.
This has compounding effects over time. BACKSPACE users build personal libraries of refined prompts that they reuse and remix. They develop intuition for which models respond well to which prompt patterns. They notice subtle behavior shifts when models update. By year three of intensive AI use, a BACKSPACE has become something approaching an expert practitioner — while a CTRL+V is still just copying and pasting.
BACKSPACE has a failure mode worth naming: over-investment in prompt craft for low-stakes tasks. If you spend 15 minutes refining a prompt to write a 3-line email, you've lost the productivity battle. The AI was supposed to save you time. You spent more time prompting than you would have spent writing.
The signs you're falling into the BACKSPACE trap:
The cure: pick your battles. Reserve the BACKSPACE treatment for outputs that actually matter. For everything else, accept "good enough."
BACKSPACE users have a real and measurable career advantage in AI-heavy workflows. They produce better output, debug AI failures faster, and develop intuitions that compound. As more jobs become "human + AI" rather than human or AI, BACKSPACE skills are increasingly the difference between mid-tier and top-tier performers.
If you're a BACKSPACE, lean into it. The world is finally rewarding the skill you've been quietly developing for years.
BACKSPACE users get the best results from JARVIS (precise output for precise input) and WATSON (deep knowledge that BACKSPACE prompts can extract). They're frustrated by BARD (too verbose), CLIPPY (no real opinions to extract), and GPT-F (hallucinations defeat the precision).
Curious if you're a BACKSPACE or just dabbling in prompt craft? The AIBT human test reveals which of 16 keyboard-key user types you actually are.
→ Take the Test