> ## Documentation Index > > Fetch the complete documentation index at: https://lithi.ai/llms.txt > > Use this file to discover all available pages before exploring further. # What if the compute we need is already on our desks? Lithi connects computing work with Macs people already own. You approve the price, and Mac owners choose when to help. [Get a free estimate](/compute) [Earn With Your Mac](/connect) ## Start with one prompt. 01. **COPY THE PROMPT** — Copy the prompt into your AI agent. It reviews the work you authorize. 02. **SEE WHAT FITS** — Your agent suggests suitable tasks and shows what they could cost. 03. **CHOOSE YOUR NEXT STEP** — You decide whether to stop with the estimate or ask your agent to help with setup. ## See how Lithi works. A business sends a batch of work. Suitable Macs process separate parts. Lithi checks the results and returns them together. [Ask Lithi a question](/) ## How can Lithi help? Find AI work you can hand off. Start with repeatable work. Your agent checks which tasks fit Lithi. - **Run AI inference** — Put thousands of prompts through the model you choose. - **Evaluate model outputs** — Check answers for accuracy before you rely on them. - **Label datasets** — Tag and sort training data with one set of rules. ### Compare AI models side by side. Ask each model the same questions. Compare the answers using one set of scoring rules. See where the models disagree. - **Choose a model** — Compare candidates on the same test set. - **Check a release** — Re-run the same tests and see what changed. - **Review disagreements** — Send disputed answers to a person for review. ### Translate your content into the languages you need. Send your content library as one batch. Lithi checks the translations and returns them organized by language. - **Translate for your markets** — Prepare one library for every language you sell in. - **Localize a series** — Prepare translations for one video or podcast series. - **Keep the labels** — Translate training sets without losing labels or structure. ### Find decisions in your meeting recordings. Send the recordings you want to review. Get a list of decisions, who’s responsible and quotes showing what was agreed. - **Find the decisions** — Turn a quarter of calls into one decision list. - **Compare interviews** — Compare fifty interviews and find where people agree. - **Keep the record** — Search what was said, with the timestamp attached. ### Turn documents into usable data. Send the documents you want to review. Get extracted information linked to the page it came from. - **Extract key fields** — Read totals, dates and names out of statements. - **Compare contracts** — Find the clause that differs across two hundred files. - **Organize a PDF library** — Turn a library of PDFs into a sortable table. ### Clean up your data before you use it. Send the messy export. Get back tidy, labeled rows and a record of what changed. - **Clean a training set** — Remove duplicates and fix formats without losing labels. - **Check a migration** — Check whether your exported data fits the new system before you move it. - **Agree on formats** — Make names, dates and units match across sources. ### Benchmark your models on real Macs. Test supported models on compatible Macs. Compare their performance. Each result includes the inputs and test conditions. - **Compare two builds** — Test two versions of a model on real hardware. - **Check each model release** — Run the same benchmark on every release. - **Compare compatible Macs** — See how one model behaves across a range of machines. ### Build and test Apple apps in parallel. Send your builds and tests together. Get the result, logs and files from each run. - **Check a release** — Collect build and test evidence on approved Macs. - **Test every branch** — Run checks for each pull request independently. - **Cover more systems** — Check one build against several macOS versions at once. ### Turn podcasts and videos into reusable content. Send a whole season at once. Get transcripts, clips and supporting text for each episode. - **Search past episodes** — Turn years of episodes into text people can find. - **Prepare chapters and clips** — Prepare chapters and cuts while the edit is fresh. - **Prepare clips and text** — Prepare clips and text from one recording for the channels you use. ### Give your AI agent sources it can cite. Send the documents your agent needs to use. Get prepared sections linked to their source pages, so your agent can cite them. - **Use your own documents** — Answer from your own documents, with sources attached. - **Prepare your archive** — Split years of wikis without losing headings or structure. - **Trace a claim** — Follow any answer back to the page behind it. ### Review repetitive code changes in bulk. Share the code you’re allowed to provide. Get findings and suggested changes to review with your agent. - **Review a large change** — Review a planned change across the code you choose to share. - **Find missing tests** — Find gaps in test coverage and review proposed checks. - **Group repeated findings** — Group repeated findings so your agent can review the changes. ## How we protect your work. Lithi keeps task content encrypted while it moves through the network. **IN TRANSIT** — Your tasks and results travel in encrypted form. **AT REST** — Stored tasks and results remain encrypted. **IN USE** — The approved Mac decrypts and reads its assigned input while processing it. Read how encryption protects your data and where its limits apply. [View Trust Center](/trust) [Read the encryption details](/trust/encryption) ## Questions ### What kinds of work can Lithi handle? {#suitable-work} Start with repeatable tasks whose results you can check. Examples include translating content, turning meeting recordings into notes and organizing data. Your agent checks whether the task and its data fit Lithi before you continue. ### Does Lithi store my task data? {#task-data} Yes. Stored task content is encrypted. Lithi also keeps the minimum records needed for billing and to show what happened. Content is readable during authorized preparation, on the Mac processing it and when you decrypt the result. A sufficiently privileged or compromised Mac operator may be able to inspect it during processing. Splitting work does not make it anonymous. Lithi’s key-release system can technically decrypt task keys. It releases them only to the assigned Mac after the required assignment and access checks, and does not use them to read your content. ### Does every job run on a Mac? {#where-work-runs} Yes. People with supported Apple silicon Macs can take part, including with one personal Mac. Owners choose when to help. Lithi’s cloud systems coordinate the work and its price. Which Mac runs a task depends on the rules for that work and the capacity available; additional personal Macs and business administration are optional paths. ### Can Lithi use Macs in my area? {#nearby-macs} Sometimes. Lithi can use approved Macs in your broader region when they’re available and meet the rules for your work. We can’t promise a particular Mac or location, and each owner chooses when their Mac takes part. ### How do I get started? {#getting-started} Copy the prompt into your AI agent. It helps you find suitable work and get a free estimate. The prompt does not install Lithi, upload your files, request API keys or start work. You choose whether to apply for access. Your agent needs your approval to send an application. Before paid work starts, you review and approve a final signed quote. ## Choose how you take part. Bring AI work to Lithi, or let your Mac help run it. [Get a free estimate](/compute) [Earn With Your Mac](/connect)