Every release,in full

What shipped, release by release. Written when the build lands, not after.

2026-08-10

Fix

Emoji and Persian text come back the way the agent wrote them

Khint strips invisible characters out of any text it handles, because they are how a hidden instruction travels inside something that looks ordinary. Two of the characters on that list are not invisible decoration at all: one is what joins a family emoji into a single glyph instead of three separate people, and the other is a letter in Persian and in the Indic scripts. Removing them from text coming in is right: that is where a hidden payload would arrive. Removing them from text going back out was corrupting every emoji sequence and every Persian word an agent produced, on every paste and inside every document Khint edited for you. Those two are left alone on the way out now; everything genuinely invisible still goes, on the paste path, the ask-in-words path and the document path alike. Then the one model answer that reached storage untouched: the running summary of your work session. It is never shown and never pasted, but it is read back into every prompt for the rest of the session, so anything an entry smuggled into it travelled with you all day. It is cleaned now, on both routes that write it. That same summary could also be scheduled for a rebuild and have the rebuild cancelled in the same breath: setting a session objective and the automatic refresh every fifteen entries were both marking it stale in a way a rebuild already under way could wipe, leaving agents reading an older picture than the one you had asked for. Three smaller ones ride along. The little note under a run saying which context was used was worked out by reading your session and your knowledge a second time, so it could describe something other than what was actually sent, and made every run pay for a second search to do it; it now reports exactly what went. Khint's connector for other AI tools now publishes how long an entry may be, instead of letting an assistant discover the ceiling by being refused with the work already done. And the "manage subscription" button in Billing used to print the payment provider's own sentence, customer reference and all, when something was wrong with the account; it says what is actually happening and what to do about it.

Emoji sequences and Persian, Hindi and other Indic text are no longer broken on the way back from an agent, pasted or written into a document.
The running summary of your work session is cleaned before it is stored, and can no longer be re-read into every later prompt carrying something hidden.
A summary rebuild can no longer be cancelled by the very thing that asked for it, so agents stop reading a stale picture.
The context note under a run describes what was really sent, and stopped charging a second search to say it.

2026-08-10

Fix

A Super Khint run that is cut short still tells you what it did

Super Khint proposes a list of actions, you run the ones you want, and when you close the card it writes down what happened: a line on your home page and an entry in your work session. There were three ways a run could end and only one of them wrote that record. Close the card while a write was still going out and the write landed with nothing left to attach its receipt to: the ticket existed in Jira, the run left no trace anywhere in Khint. Start a second Super Khint while a card with completed actions was still open and the first run's receipts were overwritten where they stood: closing that first card afterwards then did nothing at all, in silence, for actions that had genuinely written to Jira or Slack. All three endings write the record now, and whoever finishes last owns it, so closing under a slow write waits for the write rather than pulling the run out from under it. The fourth fault is the one that could cost you a duplicate ticket. The panel can rebuild itself from scratch (on Windows the system recycles it), and the rebuilt list came back with no memory of which rows had already run: every line offering a live Run button again, and one press of Run all re-creating tickets that already existed. Rows that succeeded are remembered across a rebuild, and the guard sits at the moment of execution rather than in the display, so a second press hands back the receipt it already has instead of writing a second time. And a question the panel asked you (a Jira key, a page address) that disappeared under one of those rebuilds used to take the keyboard with it, leaving you typing into your own app with the keys going somewhere else, and left Run all stuck for the next plan too. A question that goes away now answers itself with a cancel, which is what it meant.

Every way a run can end (closing the card, a write finishing late, a second run taking the slot) now leaves the same record.
Closing the panel during a write waits for the write instead of losing its receipt.
After the panel rebuilds itself, rows that already ran stay marked, so Run all cannot re-create tickets you already have.
A question that vanishes hands the keyboard back and no longer jams the next Run all.

2026-08-10

Fix

Your clipboard comes back, and a workflow shortcut works from any pack

The palette borrows your clipboard for a moment: it copies whatever you have selected so it can read it, then puts back what was there before. Every way out of the palette owes you that, and four of them had quietly stopped: Capture, the PowerPoint diagram, sending to Claude, and opening another app all left the borrowed copy parked, where a later run with no palette even open could write it over your clipboard. That was one forgotten decision per row, fixed one row at a time while the next one shipped; there is a single rule now, so the row added next is covered by default. PowerPoint is the one deliberate exception, because it reads that copy itself. Three more places had the same shape: a plan started from the ask-in-plain-words bar never restored anything at all, and Super Khint restored only when it had read your selection, never when it read your clipboard or a file, and not at all if you dismissed the file picker. Then the shortcuts. A workflow's own key combination is registered across all your packs on purpose, but pressing it only ever looked inside the pack you had active: if the workflow lived in another one the key did nothing and said nothing, while the very same workflow started by voice worked, because that route looked everywhere. Both look everywhere now, and a workflow that genuinely cannot start says so instead of failing in silence. Two more you would have felt without being able to name. On Windows, any app whose name merely contains "cmd", Total Commander among them, was having text typed into it character by character and losing the first one. And the palette opened on whichever screen it was last on rather than the one your pointer is on, so on a multi-monitor desk it could appear on a display you were not looking at. Finally, the ask-in-words bar and Super Khint were refusing runs that write to Jira or Slack unless a work session was active: a rule that stopped applying weeks ago, that no other route enforced, and that was blocking work which would have gone through perfectly well.

Every way of leaving the palette now puts your clipboard back, including the four rows that had stopped doing it.
A workflow's shortcut works wherever the workflow lives, instead of silently doing nothing outside the active pack.
On Windows, apps whose name contains "cmd", Total Commander and friends, get their text pasted properly again.
The palette opens on the screen your pointer is on, and no longer refuses Jira or Slack writes for want of an active session.

2026-08-10

Fix

What Khint files appears once on your home page, with a link that opens it

Every ticket, page, comment or message Khint sends into one of your tools leaves a line on your home page with a button to open it. Four things were wrong with that trail, and all of them concerned writes that really landed. A workflow's Jira step produced two lines: the proper one, saying what was created and linking to it, and a duplicate carrying the same text with nothing to click. Only the informative one is written now, and steps that have no line of their own (a read, or a message sent to Claude) keep theirs. A fan-out keeps its own line too, deliberately: when several branches write at once, that summary line is the only place a branch that failed is visible. Linear left no line at all. It was the only tool writing into your work session and never onto your home page, and it now uses the same path as everything else, which also means a local hiccup while writing that line can no longer make the command report failure for an issue that exists in Linear, the exact way one ticket becomes two. The "Open" button was finding its address by scanning the text for the first link in it, on the assumption that every write leaves its own. Five do not: appending to a Confluence page, commenting on one, and both of the Gmail writes all record a preview of your own content, so the button opened whatever link happened to be sitting inside the paragraph you had just appended. A newly created Confluence page had no working link at all, even though the right one had been worked out moments earlier. Links are now passed along rather than guessed, and where getting one would mean going back to the tool a second time (Slack, and the two Gmail writes), there is simply no button, because no link beats a wrong one. And the line itself was the one place text coming back from an outside tool reached storage without being cleaned first.

One home-page line per ticket or page instead of two, with the one that carries the link kept.
Linear writes now leave a trace like every other tool, and a local hiccup can no longer report failure for an issue that exists.
"Open" goes to what Khint created, never to a link that happened to be in your own text, and shows no button rather than a wrong one.
Titles and messages coming back from your tools are cleaned before they reach your history.

2026-08-10

Fix

Filing a Jira ticket stopped failing after you had already confirmed it

Before Khint files a ticket it reads your project: the issue types, the priorities, the people, and the fields your project insists on. Five faults in that circuit, and each of them showed up as a rejection that named no field, or a draft that never came back, after you had confirmed and after the earlier steps had already been charged. First: when that read failed, the failure was remembered for half an hour. A project that requires nothing and a project Khint could not read look identical in the answer, so one network blip meant thirty minutes of tickets filed without ever asking you for a required field, with no way back short of restarting the app. A failed read is still used (refusing to file a ticket that would have worked is worse): it is simply not remembered. Second: when the issue type Khint had remembered for a project no longer existed, it correctly fell back to another type but kept the rules it had read for the dead one, so the card offered no priorities, no people, and none of the new type's required fields. It re-reads for the corrected type, for free, and the text you had drafted survives. Third: an answer Jira cannot take ("about 5" in a field that wants a number) was treated as no answer at all, so the same question came back word for word until Khint gave up, and two of the places that file tickets then went ahead knowing a required field was empty. Every one of them now says why an answer will not do, counts a wrong answer separately from a missing one so a typo does not use up the form's patience, and refuses to file rather than send something that will bounce. In the workflow editor the same confusion erased what you were typing as you typed it; the box keeps your text and explains why it is not saved yet. Fourth: the drafted ticket was the one model answer that skipped Khint's cleaning step, on its way into a system Khint has no way to tidy up afterwards, and into your work session. Fifth: on a large customised project, the project's own catalogue of fields could push the request past the size the server accepts (measured 60% over on a realistic one), which failed the draft outright; the catalogue is now trimmed to fit rather than allowed to crowd out your text.

A failed read of your project is no longer remembered for half an hour, so a blip stops costing you thirty minutes of tickets missing their required fields.
A remembered issue type that no longer exists no longer leaves the card showing another type's priorities, people and fields.
An answer Jira cannot accept is explained instead of asked again, and the ticket is refused rather than filed knowing it will bounce.
Drafted tickets go through the same cleaning as every other answer, and a large project's field catalogue no longer breaks the draft.

2026-08-07

Fix

Knowledge now indexes everything it said it was indexing

Knowledge builds a local index of your Jira, Confluence, Notion and Linear so an agent can answer with your project's own material. It has been quietly indexing less than it claimed, and the failures were all silent ones: no error, no warning, a card saying everything was up to date. Entire Confluence spaces were never read. The index kept one bookmark for all of them, so a large busy space moved that bookmark far forward and every quieter space visited afterwards looked already-done and was skipped, permanently. Each space keeps its own bookmark now. Above fifty spaces the list simply stopped, so an organisation with sixty had ten that could never be indexed at all; the list is followed to the end. Notion pages edited within the same minute as the last one read were skipped forever, and are not any more. Then the part you could see and could not explain: for anyone west of London, "what moved" in your tools showed nothing, because Jira stamps its updates in your account's own timezone and Khint was comparing those against a clock in another one. Timestamps are normalised now, and rows written before today are repaired on their own. Switching a source off is supposed to erase what it indexed, and doing it while that source was mid-index put everything straight back, and those documents carried on being handed to agents for as long as any other source stayed switched on. The erase now holds whenever you press it, and a run only ever reads sources that are currently on. Two more: the "Search Knowledge" tick box on an agent you had already saved never actually saved (ticking it did nothing, unticking it did nothing either), and pressing "Sync now", or switching a source on while another was indexing, was quietly ignored until the next half-hourly pass. And an agent that matched a document containing one very long unbroken block used to fail on every single run with a raw error, with no way out but unticking the box.

Confluence spaces are each tracked on their own, so a busy one can no longer hide the quieter ones, and the list no longer stops at fifty.
"What moved" shows your Jira updates again, wherever in the world your account is set.
Switching a source off erases it even mid-index, and a run only reads the sources that are switched on.
The "Search Knowledge" tick box on a saved agent now saves, "Sync now" is never ignored, and a very long document no longer fails every run.

2026-08-07

Fix

A capture answers about the screenshot you just took

Capture starts by clearing your clipboard, so that when an image appears there Khint knows it is the one you have just drawn. That clearing step had a single character wrong in it and had never run, so if an image was already sitting on your clipboard, Khint took it as your capture the instant you pressed the shortcut: it read the previous picture, charged you for it, and left the selection crosshair on screen doing nothing while you were still choosing a region. Worse, it armed itself: a successful capture often leaves its own image on the clipboard, so the second capture of a session was reliably wrong. It is fixed, pinned by a test, and the step now reports a failure instead of swallowing it: losing that error is what let one character survive this long. Two ways a reply you had already paid for could be lost are closed with it. The window's red button closed unconditionally while Escape politely waited for a plan or an answer in flight; both gestures now follow one rule, and when the window declines to close it says why rather than appearing broken. And an answer still on its way back when you closed the window or took a new capture used to arrive anyway, surfacing at the top of your next, unrelated conversation, and being written into your work session as though it were about the new screenshot. An answer now belongs to the conversation it was asked in, or to nothing.

The clipboard is really cleared before a capture, so Khint reads the region you drew rather than the last image you copied.
That fault also made the second capture of a session reliably wrong; both are gone.
The red close button now waits for a plan or an answer in flight, exactly as Escape already did, and says so.
An answer arriving after you closed or re-captured can no longer land in the next conversation or be saved into your session.

2026-08-07

Fix

You are charged for the engine that actually answered

Khint runs on two AI providers so that a bad hour at one does not become a bad hour for you, and Quality costs three credits where Instant costs one. The price was being decided before the fallback happened: if your Quality run was served by the backup engine, you were charged the Quality rate for it anyway, for as long as the incident lasted. The rate is now read after the run, from whichever engine answered, which is what the Capture side already did. The mode picker in Settings was also naming an engine that had been retired days earlier, so it matched none of the cards on screen while your actions ran on the free default; it now reports exactly what a run will use. Two more, both about money. If you bring your own API key, drafting a Jira ticket runs on your key again: it had quietly moved onto the small courtesy allowance, which for a Lifetime buyer meant spending an allowance they had already paid to be free of. And a capture conversation was sending your work session's summary twice on every question, which you were paying for each time; it travels once now.

A Quality run that falls back to the backup engine is billed at the rate of the engine that answered.
The mode picker names the mode you are actually on instead of one that no longer exists.
Bring your own key: drafting a Jira ticket runs on your key again rather than on the free allowance.
A capture conversation no longer sends, and charges you for, your session summary twice on every question.

Want to be in the loop?

No RSS feed yet. Drop us a note and we'll email you when there's something worth a release post.

Email me when there's news Try the latest build