Changelog
What shipped and when, newest first. Nothing appears here until it is live.
The roadmap carries no dates, because a date is a guess dressed as a promise. This page is the other half of that rule. A line appears here only after the thing it describes is running in production, and every line names something you can go and check for yourself.
The site went public on 14 August 2026. The dates before it are still listed, because the engine had been reading the chain for days before there was a screen to read it on.
15 August 2026
$HOUND is on the site. The contract address is published in the homepage bar and on the $HOUND page, and it links to the pons page for that exact address. It is never shortened. A half shown address is the thing an address poisoning scam is built around, so the bar wraps to two lines rather than truncating.
Collections show the items they moved. A row on the NFT board carries the art of the items that collection recently traded. There is no collection level image anywhere on this chain, so what you are looking at is one item standing in for the set, and the board says so rather than implying a logo.
Copy and explorer on every wallet. The board shortens an address to fit its column, but the copy button always hands over all 42 characters, and the explorer link always opens the wallet the row is actually about.
Autopsy leads with 24 hours. It used to lead with everything the read reached, which is a number describing no nameable period. It now leads with the same window the board ranks on, so one wallet reads the same in both places, and the full read stays on screen in its own panel with its own label. Wallet Autopsy.
A cockpit built for a phone. Below 1024px the terminal draws a master and detail screen instead of the desktop grid, and the grid unmounts rather than re-rendering once a second behind glass nobody is looking at.
Releases stopped costing downtime. The app and the engine each deploy into an idle slot that has to answer on its own port before any traffic is pointed at it. The engine now refuses requests while its book is still filling instead of answering with half of one, and the terminal header says which state it is in. Engine.
The docs grew a roadmap, a $HOUND page, and a page on why this exists.
14 August 2026
$HOUND launched on pons at 18:12 UTC. The whole supply went to the bonding curve before trading opened, so there is no team allocation and no vesting chart to disclose. $HOUND.
The terminal went public at houndterminal.trade,
and the engine got a public door at api.houndterminal.trade. Its status
probe is open to anyone.
NFT collections joined the board. They rank beside tokens without pretending to be tokens. No sparkline, because a chart drawn through sale prices of different items is a claim about a market that does not exist, and no realized PnL, because there is no NFT ledger yet. The board prints that reason in the column rather than leaving it blank. Collections.
The wallet column went real end to end. Live activity, the wallet's last trades read back from disk, and a PnL curve that is the wallet's own equity rather than a token chart standing in for one. Panels.
The engine keeps 90 days. The market book, token identities, a daily wallet rollup and the ETH trace all survive a restart, so a reboot stopped costing two hours of warm up. Persistence.
The ETH trace became ETH. It had been drawing a memecoin under an ETH label. Fixed at the source, and the rule that came out of it is that a panel label is a claim, and the claim has to be checkable from where the data came from.
Docs at /docs, and a homepage rebuilt around them.
13 August 2026
The engine went to production on its own box. It comes back from a hard kill in under 25 seconds with its wallets rehydrated and nothing double counted, and it judges its own health once a minute and says so out loud when the verdict changes. Engine.
12 August 2026
The engine started reading money instead of venues. Coverage was measured against the chain rather than against itself and came back at 39.8%. Netting value flow per address per transaction, plus Uniswap V4's own swap event, took that to 99.7% of swap transactions detected and 97.4% of the position moving ones caught. What it still misses is counted on the health endpoint rather than rounded away. Coverage.
The wallet leaderboard. The centre of the terminal started ranking
wallets on closed positions with cost basis, instead of open positions marked
to market. Wins print as a raw fraction, 43/45, never a percentage, because
a percentage puts one lucky close above a real record.
The board.
10 August 2026
The first cockpit, on a live feed, on a laptop. Four days before anyone else could open it.