Risks and Open Problems
Status: Live
A register of the known risks to the network and of the problems its design leaves open, with the state of each on 2026-10-08, and of the price feed, the treasury and the witnesses on 2026-11-03. It changes when a risk does.
This page lists what could go wrong, how far each risk has gone, and what limits it, with a link to the page that explains it. Security vulnerabilities are handled under the Style Guide's disclosure rule: reported privately, and documented once fixed.
PXS and the economy
PXS promises no price: nothing defends it, and no one owes its holder anything. The rows below are risks in the mechanism that converts it into PIXA, and in the economy that mechanism rests on.
| Risk | What could happen | Where it stands | What limits it | Details |
|---|---|---|---|---|
| What a PXS delivers can fall | Conversions settle 3.5 days after the request, at the median less the haircut, which falls toward zero below a collateral ratio of 7/3. PXS has no market and PIXA one thin market, so a holder may find no buyer at the value the median names. | Printing on, no haircut; no orders on the chain's own PIXA/PXS order book (2026-11-03) | Nothing defends it | Haircut, Corridor and Settlement |
| The median still rests on the placeholder | Every collateral ratio and threshold distance is still computed mostly from an assumed PIXA price. The witnesses' market readings ranged from 31.731 to 70.792 PIXA per PXS; once most scheduled witnesses read the market, the median and every ratio move to their middle within about two days. | Median 51.000 PIXA per PXS; 76 of its 84 samples are placeholder values, and 7 of 21 witnesses still publish one (2026-11-03) | Nothing: the medians delay a switch by about two days but do not soften it | Oracle and Price Feed |
| One thin market for PIXA | Every market reading of PIXA comes from one order book. Real orders that move it move every such feed at once: the median across witnesses does not filter them, and only the 84-hour median delays them. | One market, PIXA/USDT on Coinstore; at 1,000 PIXA a side, the book read 0.13402 to 0.143006 USDT (2026-11-03) | The fair value cannot leave the spread; bigmac-feed stops publishing when the spread passes 25% |
Oracle and Price Feed |
| The treasury holds most PXS | PXS paid out of the treasury enters the debt ratio. If all of it were outside, the collateral ratio would be about 7.6 instead of about 155. | 248,363.319 PXS, 95.11% of all PXS (2026-11-03) | At most 1% of the fund a day, and only to proposals that stakeholders approve | Haircut, Corridor and Settlement |
| The Big Mac reference | One good stands for purchasing power. Its price differs by country, changes in steps when The Economist publishes a new edition, and depends on a third party continuing to publish. Local prices that witnesses declare cannot be checked on chain. | Each witness uses its own country's price from the July 2026 edition, or a price it declares; a later edition reaches the feed only with a new release or setting (2026-11-03) | The median across witnesses; the reference can be changed in the feed software | Oracle and Price Feed |
| The design and the chain disagree | Readers of the design notes expect a corridor of 3 to 10, 21 witnesses and a basket; the chain enforces other thresholds, elects up to 21 witnesses with 1 required, and reads one good | Thresholds unchanged since genesis; 21 witnesses scheduled on 2026-11-03 | Every PXS page states the chain's numbers | PXS at a Glance |
| Words that promise | Words that imply a peg or a stable value would invite reliance on a promise that does not exist | — | The wording rules | Style Guide |
| Sustainability of rewards and the fund | Issuance funds rewards, witnesses and the fund. If the network does not grow, it dilutes holders without drawing new participants; the fund shrinks while it pays more than it receives. | Rate 9.75% a year, issued exactly since hardfork 30 (about 11.3% before it); the fund shrinks by about 32 PXS a day while proposal 1 runs (2026-10-08) | The rate falls by 0.01 percentage points every 250,000 blocks; proposals need votes | Supply, Inflation and Yield, Decentralized Pixa Fund |
| Distribution of the allocations | Most stake has yet to reach users. How it reaches the market shapes PIXA's market value, and with it every PXS ratio. | 72.29% of stake in pixa.rex and pixa.team (2026-10-08) |
The allocation accounts cannot vote; pixa.rex is to be distributed within 9 to 11 years of the TGE, a commitment the protocol does not enforce |
Genesis and Distribution |
Governance
| Risk | What could happen | Where it stands | What limits it | Details |
|---|---|---|---|---|
| Few voters | A handful of accounts elects every witness and sets the fund's threshold | 7 accounts voted for witnesses, five of them with stake, and the five largest holders held 93.00% of the stake outside the allocations (2026-10-08) | The 30-day delay on new stake; votes expire after a year; the allocations cannot vote | Decentralization and Safeguards |
Technology
| Risk | What could happen | Where it stands | What limits it | Details |
|---|---|---|---|---|
| Dependence on Hive's code | Pixa must carry its own changes forward onto each new Hive release, or fall behind its fixes | Built on hived 1.28.7, now 1.30.0 with hardforks 29 and 30 | A short, documented list of Pixa changes | Differences from Hive, Protocol Upgrades |
| Tools that age | Libraries and services lag behind the chain or carry quirks | @pixagram/dpixa 1.5.4 needs workarounds; the public gateway rewrites some field names inside post text |
Documented workarounds | SDKs and Libraries, Differences from Hive |
| Storage that only grows | Every artwork stays in the block log for good, so a full node's disk keeps growing | Artwork posts had a median body of about 26 kB in the first month; the largest was 497,479 bytes | Resource Credits price every byte; the block size is a witness vote | Pixel Art On Chain, Run a Node |
| Content that cannot be removed | Data published by mistake, personal data, or content unlawful somewhere stays in the history every full node keeps | — | Apps and indexers can hide content; the record cannot change | Censorship Resilience and Moderation, Permanence and Provenance |
| Account recovery is new | A stolen owner key can be recovered only if it was changed after hardfork 30; an owner change made before 2026-10-07 12:00 UTC left no record to recover from | Owner history recorded since block 949,330 | Keep the owner key offline; name a recovery partner | Keys and Recovery, Chain Parameters |
| No testnet | Every test transaction is real and permanent | — | Test with a small account | Developer Quickstart |
Product and growth
| Risk | What could happen | Where it stands | What limits it | Details |
|---|---|---|---|---|
| A narrow medium | Pixel art draws a smaller audience than photography or video | 141 artworks published in the first month (2026-09-04 to 2026-10-04) | Blog posts and communities widen what members can publish | Pixel Art On Chain |
| The network effect | A young network has few readers, so few rewards, so few reasons to join | 100 accounts (condenser_api.get_account_count, 2026-10-08); since hardfork 30 any account above 2.5 Pixa Power moves rewards, where before only 21 could |
Free sign-up; stake reaches authors through rewards | Voting and Curation, Decentralization and Safeguards |
| The marketplace is not built | Selling artworks, the project's stated aim, is not yet possible on chain | Planned for 2027 | — | NFTs and Marketplace |
| Dependence on the operator | The app, the account service and free sign-up rest on one operator, Pixa Rex S.A. | Sign-up is funded by a fund proposal until 2026-12-31 | The chain does not depend on the app; anyone can run a node or another app | Who Does What, Roadmap |
Legal and regulatory
| Risk | What could happen | Where it stands | What limits it | Details |
|---|---|---|---|---|
| Changing rules on crypto-assets | Regulators may classify PIXA, Pixa Power or PXS differently from one another, and differently by jurisdiction; rules such as the EU's MiCA regulation continue to develop | These pages state no legal classification | — | The Regulatory Status page (planned) will hold the approved text; Style Guide |
Open problems
| Problem | Why it is open | Where it is discussed |
|---|---|---|
| Leaving the placeholder | The witness repository dropped the placeholder on 2026-10-30, but seven of the 21 scheduled witnesses still published it on 2026-11-03. When the rest switch, and how the network absorbs the jump in every ratio, is not decided |
Oracle and Price Feed |
| Independent readings of PIXA | The witnesses' Big Macs are independent, but every market reading reads the same order book. The software accepts several markets (--pair); PIXA trades on one |
Oracle and Price Feed |
| A basket beyond the Big Mac | The design notes name a coffee as well; nothing reads it, and nothing says who would decide | Oracle and Price Feed |
| Design and chain thresholds | Either a hardfork brings the corridor into the code, or the design adopts the chain's thresholds | Haircut, Corridor and Settlement |
| The treasury and the debt ratio | How PXS paid out by the fund should be planned against the print stop and the haircut | Haircut, Corridor and Settlement |
| Measuring the Atlas | The chain counts PIXA and values PXS against it at the median; it cannot see how much a market would absorb | The Four Components |
| What a sale settles in | Whether the marketplace settles in PIXA, PXS or both | NFTs and Marketplace |
| Who pays for history | Full nodes keep every artwork forever; nothing pays them for it directly | Run an API Node |
| Spreading stake and witnesses | The protocol separates powers; distributing stake, adding witnesses and finding voters is work outside it | Decentralization and Safeguards |
Sources
Each row links to the page that sources its figures. Figures dated 2026-10-08 were read at blocks 979,298 to 979,332 with condenser_api.get_dynamic_global_properties, condenser_api.get_feed_history, condenser_api.get_accounts, condenser_api.get_order_book and condenser_api.get_account_count; those dated 2026-10-06 at block 917,250.