Graphics acceleration

An optional pack lets the AI assistant run on a dedicated NVIDIA graphics card instead of your processor. Nothing else changes.

• If you have a dedicated NVIDIA graphics card, you can now let the AI assistant run on it instead of on your processor. It is a one-time download in Settings, under AI Model. In our testing on one machine with an NVIDIA RTX 5070, the AI produced answers about 7 times faster than it did on the processor. That is one machine and one model, so treat it as an indication rather than a promise for your hardware
• It is genuinely optional. Nothing prompts you twice and nothing installs itself. A PC that skips it behaves exactly as it did before, and you can remove it at any time to return the AI to your processor
• The pack is built on Vulkan, a graphics standard that NVIDIA, AMD, and Intel hardware all support. This release only detects NVIDIA cards, so only NVIDIA users are offered it today. Widening that to AMD and Intel is a planned update to the detection step and will not need a new download
• An honest limitation: graphics acceleration stays off whenever SimpleFixAI is started as administrator. That is a deliberate safety boundary, not an oversight. The pack lives in your personal user folder, and loading it into a process with administrator rights would be a real security risk, so the app says so plainly instead of pretending no graphics card was found. Start the app normally to use it, and a future release will move the pack to a protected location and remove the restriction
• Every file in the pack is checked against fingerprints built into the app before each use, so a tampered or damaged file is refused and the AI simply continues on your processor. If a graphics attempt ever fails, the app returns to your processor so the AI keeps working. Nothing about repairs, diagnosis, or undo changed in this release