Replit

Where it stands

Built only from what we can measure or quote without using the tool. How ranking works

Read automatically from Replit's public website: how reliable, secure, private and open the site is. This is not a test of how well the tool does its job. How this is measured

• Reliability5 / 5The page answered and finished downloading in 65 ms.
• Security5 / 54 of 5 protective headers set (strict-transport-security, content-security-policy, x-content-type-options, x-frame-options).
• Privacy5 / 5No third-party tracking services found on the page.
• Transparency3 / 5No privacy policy linked; links a pricing page.

92% of the 12 people whose comments we could classify were positive. Read automatically from public Hacker News threads about Replit; the classifier is simple and can misread sarcasm, so check the threads. How this works

• Positive

  “The former two, though, run the code entirely on-device, but from the looks of it Replit will run code on their servers. While that's not a surprise (given repl.it's model), it's definitely a little less useful for me…”

  Read in the thread
• Positive

  “The demos I've seen of Replit Ghostwriter indicate that it is still possible to get good results without FIM, as long as you have good enough software around the model, but I think FIM could still improve things further.…”

  Read in the thread
• Critical

  “> When fine-tuned on public Replit user code, the model outperforms models of much larger size such as CodeLlama7B: The table just below this shows the other models doing better on half of the benchmarks; the Replit…”

  Read in the thread

Threads: Replit Mobile App (2022, 207 comments) · Replit now supports every programming language in Nix (2021, 81 comments) · Replit's new AI Model now available on Hugging Face (2023, 51 comments)