Ep 197 Thread 1:12 w/ Justy

LLMs process text from left to right — each token can only look back at what came before it, never forward. This mean...

Repeat your prompt to boost LLM accuracy without extra cost.

Embed this episode

Paste this on any site — the player is a self-contained iframe with no cookies or trackers.

<iframe src="https://sandrise.io/exploring-next/embed/197"
  width="100%" height="180" style="max-width:640px;border:0;border-radius:12px;overflow:hidden"
  title="Exploring Next — Episode 197 audio player"
  loading="lazy" allow="autoplay" referrerpolicy="strict-origin-when-cross-origin"></iframe>
Embed & API docs →
Script GPT-OSS 20B Voice ElevenLabs v3

Transcript

Justy LLMs read text token by token from left to right. That means the first word can only see what came before it, never what comes after.

Justy So when you give a long prompt with background at the start and a question at the end, the model answers knowing the context, but the context never saw the question.

Justy Researchers tried a simple tweak: send the same prompt twice in a row so every token gets a second pass that can attend to the whole input, for better clarity.

Justy The result? Accuracy jumps on seven benchmarks across Gemini, ChatGPT, Claude, and DeepSeek. No extra output length — just the hardware's parallelism at work without overhead and keep latency low.

Justy They found that sometimes the boost is modest, but sometimes it's huge: one model went from 21% to 97% on a name-search task.

Justy No extra loss, no finetuning, no clever prompt engineering needed. Just duplicate the prompt and let the model run its course every time.

Justy So if you want tighter answers without paying for longer outputs, copy your prompt, paste it again, and give the model a second look today.

Justy Give it a try today—double your prompt, watch accuracy surge in minutes.