> For the complete documentation index, see [llms.txt](https://docs.thetpa.ai/deai-tokenized-processing-architecture/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.thetpa.ai/deai-tokenized-processing-architecture/tpai-chain-full-architecture/open-items-for-the-tech-team.md).

# Open items for the tech team

These details are not decided. The tech team proposes them, and the founder approves. \[Open]

| Item                           | What is needed                                                                                                                                                                                                                                                                            |
| ------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| TPE specification              | Language, determinism rules, module interfaces, how AI contracts call each other                                                                                                                                                                                                          |
| Compute unit                   | The exact unit for pricing and metering (suggested: a normalized GPU-second)                                                                                                                                                                                                              |
| Reward weights and chain bonus | The formula that makes each tier earn more than its stake grows; the micro-node income floor                                                                                                                                                                                              |
| PING priority                  | How priority is calculated; the value of the fair-share cap                                                                                                                                                                                                                               |
| Cost table                     | Update timing and the utilization factor per GPU type                                                                                                                                                                                                                                     |
| Community & Ecosystem budget   | One model covering staking rewards, PING rewards and day-one validator income until vesting starts on 11 October 2028: yearly spend for each, how it changes with TPA sales and the Meta Energy Level, how long the pool lasts, and node income per tier until AI fees cover server costs |
| License and upgrade fees       | Where the one-time fees go                                                                                                                                                                                                                                                                |
| Two-step transfers             | Whether pending funds expire if the sender never acts                                                                                                                                                                                                                                     |
| dApps must use AI              | How TPE checks that a dApp really calls AI, at deploy time and while it runs                                                                                                                                                                                                              |
| dApp state                     | How much state a dApp can keep on chain, given that user data never goes on chain                                                                                                                                                                                                         |
| Overlay network at scale       | 16.7M TPA endpoints need many Kubernetes clusters joined into one overlay network; design and prove it, and plan IPv6 for growth, including how one flat /8 network is routed across every datacenter                                                                                     |
| Address formula                | The exact TPA address to IP mapping, and the cooling-off period before a freed address is reused                                                                                                                                                                                          |
| Per-holder schedules           | Loading each early adopter's SAFT lock and release schedule at genesis                                                                                                                                                                                                                    |
| Milestones                     | Format of published milestones and the validator vote that confirms them                                                                                                                                                                                                                  |
| Price check                    | Cost per call for the top 10 models at the floor price, compared with leading AI APIs, before launch                                                                                                                                                                                      |
| PING compute priority          | How the GPU scheduler applies the Meta Energy formula to compute priority in practice (queue order, batching, limits)                                                                                                                                                                     |
| PING legal review              | Counsel reviews the referral rewards before any public wording                                                                                                                                                                                                                            |
| Node licensing legal review    | Securities counsel reviews node licenses, stakes and staking rewards before this design goes further                                                                                                                                                                                      |
| Dated build plan               | Devnet, internal testnet and public testnet dates, team size, and the minimum Phase 1 must include to launch on 11 October 2027 (due mid-December 2026)                                                                                                                                   |
| Spot-check method              | Deterministic re-run settings and the probability threshold per model type                                                                                                                                                                                                                |
| Canadian data residency        | How a TPA owner chooses to keep data and compute on Canadian GPUs (Stage's Alberta and Quebec sites or approved Canadian partners), and how the chain proves it                                                                                                                           |
| Developer tools                | SDK languages, templates and a test environment for TPA owners building on TPE in year one                                                                                                                                                                                                |
| Lapsed TPAs                    | Grace period, hold period and recovery fee for an unpaid TPA, and how its fine-tuned models, agents and data are kept encrypted during the hold and deleted at the end                                                                                                                    |
| Verification parameters        | Spot-check sample sizes, challenge window length and slashing amounts, set after testnet                                                                                                                                                                                                  |
| Committee size                 | Committee size and randomness source, from the testnet benchmarks at 100, 300 and 1,000 nodes                                                                                                                                                                                             |
| Lock contract                  | Audit and published address of the pre-mainnet lock contract before 29 March 2027                                                                                                                                                                                                         |
| Node terms document            | License prices, stake amounts and reward weights per tier, published before the node sale                                                                                                                                                                                                 |
| Session balances               | Settlement interval, metering unit and dispute handling for pre-paid sessions                                                                                                                                                                                                             |
