About Ask AI

Ask AI offers another way to explore the technical advice preserved in What the Tech. It draws on archived posts and links back to the specific replies behind its answer. It is not a reply from the original volunteers or a replacement for current, case-specific help.

How it works

On selected support threads, you can request an answer about the discussion and ask one question of your own. Ask AI uses a selection of posts from the beginning, middle, and end of the thread. Long posts may be shortened, so it does not read every detail of every log.

For malware-removal discussions, the answer aims to restate the issue, explain the reported outcome, and offer relevant guidance for a similar problem today. If the selected posts do not confirm a resolution, the answer should say so. It should not reproduce old cleanup instructions.

Either kind of answer may also draw on up to three related archived topics. The citations take you to the supporting posts; the additional discussions below the answer offer more of the archive to explore.

During one page visit, you can use each answer option once, in either order. The second answer can use the first exchange as context. The transcript stays only in page memory (RAM); reloading clears it and resets the controls. There is no persistent browser quota or saved conversation history.

The daily allowance is 30 accepted requests per network address and 100 provider attempts across the site, resetting at midnight UTC. Accepted failed requests and shared cache hits count toward the network allowance. Each generation attempt, including a validation retry, counts toward the provider allowance; cache hits need no provider attempt. A short-term rate limit also applies. People sharing a network address share its daily allowance.

Limitations

AI can make mistakes. Check the cited posts. Archived advice can be out-of-date, and a citation does not guarantee that an answer is correct. The original discussion may be incomplete or may never have confirmed a fix.

Ask AI does not browse the live web or inspect your device. Guidance beyond the supplied posts is marked "General guidance" and can make up the whole answer when the archive does not support the specific question. It comes from model knowledge, not a live check. Verify current software, drivers, and vendor instructions through official sources.

Some archived malware-removal and repair steps were intended only for use under a trained helper's supervision. Do not repeat those steps or run an archived attachment as a self-service fix. Generated answers are not reviewed by the forum's former support team.

Questions and privacy

Don't include your name, email address, passwords, or other personal information in a question. New answer requests, including shared cache hits, are verified by Cloudflare Turnstile. Your question and selected public posts are sent to OpenAI when an answer needs generating. A follow-up may also include your previous question and AI answer, using a signed context receipt that expires after 24 hours. This receipt is held only in page memory and authorizes context from the same topic; it is not a browser identity.

Cloudflare Turnstile may use browser storage for verification; its processing is separate from the temporary conversation state described here. Cloudflare processes network addresses and browser/device signals to distinguish people from bots and also to improve its bot-detection capabilities. Its published purpose is security, not identifying, profiling, or targeting people for advertising. See the Turnstile Privacy Addendum and Cloudflare's privacy policy.

Questions and answers are cached for up to 355 days from generation so repeated requests can reuse an answer. Reading an answer does not extend its lifetime. Model, prompt, or archive updates may invalidate cached answers sooner. Expired records are removed by the next daily cleanup. The archive's AI client does not create cookies or store conversation state in localStorage or sessionStorage. It does not restore a saved conversation after reload. Answer and feedback requests omit browser credentials. The shared answer cache is separate from the page's temporary transcript. OpenAI's handling and retention of submitted data are governed by its service terms and the provider account settings, separately from this archive's cache.

Feedback

Helpful/not-helpful votes and clicks on cited posts or related discussions are counted without personal identifiers. The archive keeps daily totals for 90 days, not questions or transcripts in those totals. Short-lived signed receipts prevent duplicate feedback for an answer and expire after 24 hours; their duplicate-check records are removed by daily cleanup.

Daily request counters use pseudonymous, secret-keyed network-address hashes that change each UTC day. Counters are retained for seven days; raw network addresses are not stored in the AI database. Expired quota records are removed by daily cleanup. These are anonymous-access limits, not a way to identify a person across browsers or networks. Hosting-provider access logs and necessary security cookies are separate. See Cloudflare's security-cookie documentation. This is not a claim that the whole site is cookie-free.

See the privacy page for the archive's broader privacy policy and content-removal requests.