Corrections
This page lists every change we made because something published here was wrong, or silent about something a reader needed to know. Routine data refreshes and copy edits are not corrections and are not listed; the two dates at the foot of every page cover those. Each entry stays here permanently.
Policy
- A correction changes the page, its stamp, and this log — never just the page. The corrected page keeps its original verification date until a human re-verifies the facts; the "updated" date moves because the text changed. We do not bump a verification date to make a page look fresh.
- Fixes are made at the source. When a wrong figure came from a data file or a script, the file or script is fixed so the error cannot regenerate, and a check is added where one was missing.
- Absence is not evidence. If a search or export did not show something, we say "none found in the published export", not "does not exist".
- Report an error to [email protected]. We check the claim against the primary source, correct the page, and add the entry below with the date.
Log
-
What was wrong. The page said Stage 6 points "reset at the start of each new epoch" and answered "When do points reset?" with "Weekly"; Aster’s docs describe points calculated weekly that "accumulate toward your $ASTER airdrop share at the conclusion of Stage 6", with no reset. It called the team boost and referral points "two multipliers", though referral points are added on top, and described Aster’s fee referral program as paying out "in trading-fee discounts" when it pays the referrer a commission. Its comparison said Hyperliquid’s points "converted to a token", which Hyperliquid’s docs do not state, and that Lighter’s programme "is still running", which Lighter’s own docs contradict.
What changed. Re-read against Aster’s Stage 6, team-boost, referral-bonus, referral-program and tokenomics docs on 2026-10-09. The page now describes weekly calculation without a reset, separates the multiplying team boost (with its 1.05×–1.20× table and the docs’ two different entry thresholds) from the additive referral points (taken from base points excluding team boosts), adds the undisclosed-weighting, farmed-points and market-maker terms, describes the referral commission correctly, and states Hyperliquid’s and Lighter’s statuses as their docs and the tracker do. The tracker’s Aster row was re-verified the same day.
-
What was wrong. The page told readers that higher leverage means a smaller price move to liquidation and that "this relationship holds across every asset". Hyperliquid’s liquidations docs say "The actual liquidation price is independent on the leverage set for cross margin positions"; only isolated positions work that way. It also said Hyperliquid’s docs "don’t publish a single global leverage cap", though they put max leverage at 3x to 40x and publish margin tiers that lower it for large positions, and it measured the 10% withdrawal floor against the one position rather than, as the docs say, "the total notional position value of all open positions".
What changed. Re-read against Hyperliquid’s margining, liquidations, margin-tiers, contract-specifications, portfolio-margin and funding docs on 2026-10-08. The page now gives the docs’ liquidation-price formula with a worked example showing cross and isolated behaviour, the 3x–40x range with BTC’s and ETH’s tiers, maintenance margin per tier, the withdrawal floor as the docs define it, a summary of what happens at liquidation, and strict isolated as a property of some assets rather than a mode you pick.
-
What was wrong. The page said staking GMX "earns a share of protocol fees", without saying that GMX’s docs mark staking rewards as suspended, with bought-back GMX held in the treasury until a price threshold. It said leverage caps were "100x on majors, lower on commodities"; GMX’s TradFi page lists gold, silver and both oils at 100x in on-hours and 25x off-hours. It also described Express mode as needing no per-trade gas token (its fee is paid in a gas payment token such as USDC), and quoted swaps as "routed through GM pools", a line no longer in the docs, which now add KyberSwap routing for eligible market swaps.
What changed. Re-read against GMX’s trading, fees, TradFi, tokenomics and liquidity docs on 2026-10-08: the Earn section quotes the suspension and the documented distribution trigger, the TradFi paragraph gives the on-hours and off-hours caps per market and the off-hours liquidation-factor change, Express is described by its relay and gas payment token, swap routing quotes the current docs, and two quotes were made verbatim.
-
What was wrong. The page attributed "Season 1" and an "L1 season" to Hyperliquid’s docs and answered "Was there really a Season 2?" with yes; the docs describe a "points program" and an "L1 phase" and number neither. Its examples were also out of date: it listed Lighter as "points accruing, no token yet" (LIT has been live for months) and named Aster and Lighter as the only covered venues with documented points programmes.
What changed. Re-read against Hyperliquid’s points and fees docs on 2026-10-07: the history now uses the docs’ own names and adds the May 1–28, 2024 multiplier, the cross-venue examples follow the points & airdrop tracker, and the tracker row drops the season label and states the assistance fund’s fee-to-HYPE conversion and burn as the docs word it.
-
What was wrong. The guide said Aster’s docs do not define the Chase order (they now document it in full), stated BTCUSDT’s 200x cap as current although it came from a July screenshot, and quoted four lines that are no longer in Aster’s docs: the margin switch "only applies to the selected contract", "if your margin ratio hits 100%, your positions will be liquidated", a settlement fee applying "only to winning positions", and Rewards as "your gateway to the $ASTER airdrop" with "Rh Points" and a collateral bonus.
What changed. Re-read against Aster’s docs on 2026-10-07: the Chase order is described from its docs page, the leverage cap is dated to our July capture beside the docs’ "Each asset has its own maximum leverage limit", the liquidation, settlement-fee and Stage 6 points lines quote the current docs, the unfound margin quote was removed, and Shield Mode is noted as a separate product.
-
What was wrong. The trading screen guide said Lighter’s docs "don’t publish the exact discount or kickback percentages" for referrals; the docs have published the programme since 28 July 2026, and our referral page already said so. Several Lighter pages called the Standard tier free with no exception, but Lighter’s API docs say "For standard accounts only, TWAP orders charge a 1bps fee." The fees page gave Standard a 200 ms maker latency (the docs now say 0 ms) and the Plus tier 8,000 transaction and 120,000 read requests per minute (now 4,000 and 24,000). Two pages said tier switches need no open orders or positions and run once every 24 hours; the API docs limit only downgrades to once every 24 hours. A quoted line about creating an account by deposit could no longer be found in Lighter’s docs.
What changed. Re-read against docs.lighter.xyz and apidocs.lighter.xyz on 2026-10-07: the TWAP fee is stated on the guide, fees page, order types table and fee comparison; latency, rate limits and the switching rule follow the current docs; the account-creation line now quotes the API docs ("created automatically on the first deposit"); and the guide links our referral page for the published terms. The guide also adds the docs’ unified-account definition, the public pools’ no-lock-up rule and LLP access being limited to LIT stakers.
-
What was wrong. Lighter vs Extended said Extended "lists markets Lighter does not have at all" and is "the only one of the two" listing FX, indices and equities, and that this site had not verified non-crypto markets on Lighter; Lighter’s docs describe real-world-asset perps across equities, indexes, commodities and FX, which our Lighter RWA page had covered since September. Aster vs Extended likewise called Extended "the only one of the two" with non-crypto perps, though Aster lists US, Hong Kong and Korean stock perps. Lighter vs Paradex called Lighter’s Premium maker rate "level with" Paradex’s flat Pro maker fee; at 0.0040% against 0.003% it is higher. Lighter vs dYdX said Lighter’s access rules were not stated (its Terms exclude the US, Canada, the UK and China), neither dYdX comparison mentioned that governance retired most of dYdX’s markets during 2026, and Lighter vs edgeX said EDGE’s buyback-burn was "more than Lighter’s docs say about LIT", though Lighter’s docs state that LIT is bought back from trading-fee revenue.
What changed. Each sentence now follows this site’s Lighter, Aster and dYdX pages and the fee comparison: the Extended comparisons turn on FX, collateral and leverage rather than on which venue lists TradFi at all, the Paradex comparison states the maker gap, the dYdX comparisons note the wind-down, and an unsourced "no fiat on-ramp" line was dropped from Lighter vs Pacifica. All twelve July-dated Lighter and Aster comparisons were re-checked against their profiles and now carry the older of their two profile verification dates.
-
What was wrong. The guide said Hyperliquid’s oracle price "deliberately doesn’t depend on Hyperliquid’s own market data", following the docs’ price-indices page. Hyperliquid’s more specific oracle page lists its own spot mid price among the weighted-median inputs, excluding it only for assets whose main spot market is elsewhere (such as BTC) and using it alone for assets like HYPE until external liquidity is sufficient; the guide now gives the oracle page’s version and notes the conflict. The guide also said TWAP suborders go out "every 30 seconds" (the docs now say intervals are a minimum of 30 seconds, scaled to size and running time), called the unified account "the default" (the docs say "recommended for most users"), and said the testnet faucet pays out "once" (the docs do not say so).
What changed. Re-verified against Hyperliquid’s docs on 2026-10-07: the oracle, TWAP, account-mode and faucet passages now follow the docs, a quoted line on isolated liquidation was made verbatim, and the Trailing Stop order type the docs now list was added with a note that it was not in the menu when our screenshots were taken in July.
-
What was wrong. The roundup said that in May 2026 Kalshi "became the first CFTC-approved venue to offer perpetual futures in the US" and called its availability "US-first", telling readers to "check your own region before signing up". Bitnomial had self-certified a BTC perpetual in April 2025; what the CFTC did on 29 May 2026 was approve Kalshi’s BTCPERP after Commission review. And Kalshi’s perpetuals are open only to US-based users who pass KYC and a margin-account application; its referral incentives are US-only as well. The roundup also recommended dYdX to anyone weighing "platform longevity" without saying governance retired most of its markets during 2026.
What changed. The Kalshi section now quotes the CFTC release, credits Bitnomial’s earlier self-certified perpetual, states the US-only access and referral rules from Kalshi’s help centre, and describes the referral credit as the sign-up page and FAQ do. The dYdX entry now carries the 2026 wind-down (78 active against 218 in final settlement on 16 September) and its 50% trading rebate. Fees, terms and restrictions for the five venues were re-checked against each venue’s docs; dates are on the page.
-
What was wrong. Verified in July, the page said binary price markets were "the form live today", told readers to expect binary markets before multi-outcome ones, and said the official docs described only protocol-deployed recurring outcomes, so how non-price questions resolve was undocumented. By 7 October 2026 Hyperliquid’s docs described template-based third-party deployment (labelled "Testnet-only"), and its mainnet outcomeMeta endpoint listed 29 questions — sports, an award and the Federal Reserve’s October decision among them — from three deployer venues, each settled by its deployer rather than by the protocol’s price feed.
What changed. The page now separates the protocol’s recurring price markets (daily binaries on BTC, ETH, SOL and HYPE plus a daily BTC multi-price question), which settle automatically, from third-party markets, which are settled by their deployer under its own rules, and quotes trade[XYZ]’s Terms on what Hyperliquid’s acceptance of a settlement does and does not confirm. Docs re-read and the endpoint queried on 2026-10-07.
-
What was wrong. These pages described edgeX’s original StarkEx deployment (V1) as the running system, and listed the docs’ line that "edgeX V1 will cover the gas costs on behalf of users" as a current edgeX strength; on Hyperliquid vs edgeX it decided the gas comparison in edgeX’s favour. L2BEAT records that trading moved to edgeX V2 in May 2026 and that the V1 contract froze on 2026-08-14, which it calls a planned sunset. Until 2026-10-07 the status notice also presented that deployment’s upgrade risk as edgeX’s current custody risk, and the notice that replaced it cited the legacy V1 API as edgeX’s public API.
What changed. V2 is now presented as the live venue and V1 as frozen. The gas line is quoted as a statement about V1; the V2 architecture page says only that bulk settlement spares users "high mainnet gas costs for every trade execution", so no comparison credits edgeX with covering gas. The status notice cites the live V2 market API. Checked against L2BEAT, an on-chain read of the V1 contract, and edgeX’s docs and APIs on 2026-10-07.
-
What was wrong. The comparison said OLP deposits "remain paused" with the restitution plan for affected depositors "still pending", and the status notice’s headline read "OLP restitution pending". Ostium had published a two-stage recovery plan on 2026-09-30, per The Defiant, and the deposit statement had not been re-checked since our July review.
What changed. The status notice now sets out the published plan — wallets that lost $1,000 or less are repaid in full, while the 345 larger wallets choose between a $1,000 payment and a pro-rata share of a second-stage recovery pool — and the comparison summarises it and says that whether new OLP deposits are open has not been confirmed.
-
What was wrong. The page said "Trading collateral on Hyperliquid is USDC" and limited HYPE to staking and fee discounts. Hyperliquid’s portfolio-margin mode counts "all eligible assets, which are currently HYPE, BTC, USDC, USDT", with HYPE at an LTV of 0.65, and several builder-deployed perp DEXs margin in USDH, USDE or USDT0 rather than USDC. The page also left out two documented roles: gas on the HyperEVM and the 500k HYPE stake a HIP-3 deployer must hold.
What changed. Re-verified against Hyperliquid’s docs and its public API on 2026-10-06: the page now covers portfolio-margin collateral (with its eligibility thresholds and caps), HyperEVM gas, the HIP-3 deployer stake and the permanent staking-link warning, and sources the November 2024 genesis to the Hyper Foundation and The Block instead of "as widely reported".
-
What was wrong. The review said Lighter’s Season 2 points programme "is ongoing and distributes points weekly". Lighter’s market-makers page says "Points Season 2 ended on December 26th 2025" while its overview and retail pages still describe weekly distributions, so the docs do not settle it; our points page and tracker already said so, the review did not. The points page also warned that "multi-account behavior" risks disqualification, but the retail page says "We allow up to 10 accounts per user without penalty".
What changed. Both pages now describe the contradiction in Lighter’s docs instead of asserting either reading, the points page quotes the 10-account allowance, and both now cover the separate "Lighter on Robinhood Chain Points" programme that Lighter’s docs added (terms effective 10 August 2026). Re-read against Lighter’s docs on 2026-10-06.
-
What was wrong. These pages matched Variational’s docs as we read them on 2026-09-22: the token "is not yet live", "approximately 50% of the tokens" were planned for the community, weekly point distributions "will conclude no later than the end of Q3 2026", and nothing said what points become, so the tracker told readers to treat points as unconvertible. The docs then changed and our pages did not. By 2026-10-06 the $VAR page said the token "will be launched in Q4 2026" and put 32% of supply in a Genesis Distribution of "Tokens airdropped proportionally to points holdings", and the points page moved the end of distributions to Q4 2026.
What changed. The review (spec card, FAQ, points section and risk list), the tracker row, both comparisons and the llms.txt summary now quote the current token and points pages. The tracker lists the 32% Genesis Distribution as the published allocation, and every surface notes that neither page gives a launch day or a total points count, so what one point is worth in VAR cannot be computed yet.
-
What was wrong. The section, verified against Lighter’s docs on 2026-07-30, said a separate “Experimental Liquidity Provider” (XLP) pool supplied prelaunch liquidity, that pre-IPO markets still charge liquidation fees and are settled or price-adjusted after the IPO with a day’s notice, that US equity indices track futures 23/5 with a daily 5 PM ET realignment and a syntheticSpotInfo endpoint, and that cross margin was restricted on some RWA markets. Lighter’s current docs say none of this: XLP “previously served” RWA markets and LLP now does, pre-IPO markets convert to normal markets once a stable post-IPO price feed exists, the index-futures mechanism is no longer documented, and all RWA markets support cross and isolated margin.
What changed. The section and the feature-comparison cell were rewritten from the docs read on 2026-10-06; the section now links the dedicated prelaunch, pre-IPO and RWA pages verified on 2026-09-29. Found while building those pages, which were written from the current docs.
-
What was wrong. dYdX’s governance moved 21 markets to final settlement in early September — its proposal set 2026-09-04 as the delisting date, and our refresh found the smaller book by 2026-09-07. Until 2026-09-16 the site kept publishing dYdX’s coverage from a snapshot taken on 2026-09-02: the Aster and Lighter comparisons said dYdX listed 99 perp markets when it listed 78, 21 market pages still counted dYdX among the venues listing an asset that was in settlement there, and four of those pages (BLAST, GRT, PI, TSLAX) existed only because dYdX had listed the asset. The weekly refresh saw the drop on 2026-09-07 and 2026-09-14 and refused to publish it — rightly, since a venue losing a fifth of its markets between runs is something a human should read first — but the daily health run that carries such alarms had been failing since 2026-08-10 on overdue venue-status reviews, so the snapshot-staleness warning it raised from 2026-09-13 was one more line in a run nobody was reading. The comparison and index pages did show the snapshot’s date.
What changed. The dYdX floor was lowered with the settlement evidence recorded beside it, the snapshot refreshed (99 to 78 dYdX entries; the four dYdX-only pages now redirect to the index), and a standing notice on the dYdX pages explains the 2026 wind-down: 78 active against 218 settled on the venue’s own indexer on 2026-09-16, across eleven governance batches. The overdue reviews that had kept the health run red were cleared the same day, so its alarms are visible again.
-
What was wrong. Since 2026-07-17 the site said Paradex’s Pro-tier rates were not published. The trading-fees page read on 2026-09-10 publishes the full schedule — a flat 0.003% maker fee (stated as effective July 3) and a taker ladder from 0.045% to 0.035% by 14-day volume with a 0.0175% floor. Whether the schedule was already on that page on 2026-07-17 is not known; the statement was in any case stale by the time it was repeated on the Lighter and Aster comparisons on 2026-09-09.
What changed. The fee comparison, the compare-profiles data and the Hyperliquid, Lighter and Aster comparisons now carry the published ladder with its verification date; the new Paradex fees page reproduces the table. Found while building the Paradex hub.
-
What was wrong. Since 2026-07-22 the site read the rate from Lighter’s funding-rates endpoint as an hourly figure. It is an eight-hour figure: the same endpoint reports Hyperliquid at exactly eight times Hyperliquid’s own hourly rate across 95 coins, and Lighter’s docs define the hourly payment as the eight-hour rate divided by eight. Lighter’s per-day carry on the earning comparison, the earning scenarios and the BTC cross-venue table was therefore overstated eight times.
What changed. The rate is now divided by eight before use everywhere, labelled as such, and the daily live-API check asserts the eight-times relationship against Hyperliquid so a change in the endpoint’s basis fails the run instead of silently changing the figures. Found while building the cross-venue funding screener.
-
What was wrong. Both pages said the total allocation reserved for the airdrop was unstated. Aster’s tokenomics have published one since the September 2025 token launch: 53.5% of $ASTER supply sits in an "Airdrop" category released over roughly 80 months. What remains unpublished is narrower — how much of that Stage 6 receives, the points-to-token ratio, and the stage’s end date.
What changed. Both sentences now say exactly that, with the tokenomics figure and its scope. Found while building the points and airdrop tracker, whose Aster row carries the same wording.
-
What was wrong. The scheduled data-refresh pipeline had been stalled for five weeks without anyone noticing. The market index was serving ten pages for markets the venues had already delisted (ACX, ASTR, BEAM, DRIFT, IR, STORJ, VANRY, VINE, WOJAK, XDC) and was missing fifty-five real ones; the venue-coverage snapshot was 34 days old, the Hyperliquid snapshot 44 days, the funding history 30 days.
What changed. All three snapshots refreshed (604 to 649 assets), the ten delisted slugs redirect to the index, and a staleness guard now fails the daily health run whenever any snapshot is older than ten days.
-
What was wrong. The two funding charts were each auto-scaled to their own range while the caption said they were "drawn to the same scale". Hyperliquid’s axis spanned about ±0.0013% and dYdX’s about ±0.0033%, so comparing bar heights was meaningless on the one element built for comparison.
What changed. Both charts now share one axis (the larger of the two extents), each axis is labelled "shared", and a test pins the fix so a calm and a volatile series can no longer draw identical heights.
-
What was wrong. The per-venue table said GMX documented no MCP server or agent tooling. That was a false absence: the check had searched each venue’s documentation index by title only, and GMX describes an MCP server (announced as under development) and agent skills inside a page body the index does not surface.
What changed. GMX row corrected against its full documentation export. Rows for venues where nothing was found now read "none found in the published export" rather than "not in its docs", because an export is not proof of absence.
-
What was wrong. The Lighter pages, the Hyperliquid vs Lighter comparison and the roundup described LIT as a pre-launch token. LIT was already live: it traded as a market and staking paid a stated 6% APR with a three-day unstake lock.
What changed. Every "pre-TGE" and "not launched" claim swept from the site and replaced with the verified status; the 200k points-per-week figure was re-checked and kept.
-
What was wrong. The comparison shipped on 2026-07-17 with a verified stamp but said nothing about the 2026-07-15 oracle-signer-key exploit that drained roughly $18M USDC from Ostium’s OLP vault and paused all trading.
What changed. Added a dated incident notice, a trading-status row, an FAQ entry and a caveat on the verdict, sourced to Ostium’s own statement and independent reporting. Ostium’s status has since been tracked through the venue-status notice on the page.
-
What was wrong. The review stated that Aster had made no commitment to an ASTER airdrop for points holders. Aster’s own Stage 6 documentation commits to one.
What changed. Section rewritten to the documented commitment; the dedicated points page was published the same day.
How the dates on each page work, and where every dataset comes from, is on the data definitions page. Sourcing rules are in the methodology.