{"id":554,"date":"2026-11-08T18:56:18","date_gmt":"2026-11-08T18:56:18","guid":{"rendered":"https:\/\/alandix.com\/labs\/?page_id=554"},"modified":"2026-11-15T17:49:46","modified_gmt":"2026-11-15T17:49:46","slug":"oncue","status":"publish","type":"page","link":"https:\/\/alandix.com\/labs\/oncue\/","title":{"rendered":"onCue"},"content":{"rendered":"\n<p><em>an intelligent internet interface from the dot com years<\/em><\/p>\n\n\n\n<div class=\"wp-block-buttons is-layout-flex wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button is-style-fill\"><a class=\"wp-block-button__link has-black-color has-text-color has-background has-medium-font-size has-custom-font-size wp-element-button\" href=\"https:\/\/aqtive.net\/demo\/onCue\/\" style=\"border-radius:25px;background-color:#bcaed0\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>try DEMO now!<\/strong><\/a><\/div>\n<\/div>\n\n\n\n<p><\/p>\n\n\n\n<p>onCue is referenced in many papers and also used as a case study in <a href=\"https:\/\/alandix.com\/aibook\/\" target=\"_blank\" rel=\"noreferrer noopener\">Artificial Intelligence \u2013 Humans at the Heart of Algorithms<\/a> (2ed, Chapman and Hall, 2025) and <a href=\"https:\/\/alandix.com\/ai4hci\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI for Human\u2013Computer Interaction<\/a> (CRC Press, 2027). The  web demo (above) has been built in order to give a feel for what this was once like, and this page describes more of the design principles underlaying it.<\/p>\n\n\n\n<p>Back in 1998, just before the dot-com frenzy began in the UK, Russell Beale, Andy Wood and I founded a company called aQtive, funded principally by 3i.  This was a bit of a coup as 3i made very large investments, that were on the conservative side of &#8216;venture&#8217;, such as management buy-outs.  We were completely off radar, but perhaps the fund manager we dealt with had an eye to the growing internet developments over the Atlantic at the time.<\/p>\n\n\n\n<p>Our principal product was onCue, an intelligent internet interface based on data detector technology and highly flexible architecture based around components that we called &#8216;Qbits&#8217; \u2013 we had a lot of fun thinking of &#8216;Q&#8217; names, and the name <a href=\"https:\/\/alandix.com\/labs\/calq\/\" data-type=\"page\" data-id=\"318\">calQ<\/a> came from this time, although we never developed it then.<\/p>\n\n\n\n<p>onCue was a small desktop application that presented itself as a small palette of icons at the side of the screen.  onCue watched the system clipboard and would try to work out what kind of contents were there.  It would then adjust the icons in the palette (called services) to web and desktop resources that might be relevant and giving one click access to them, for example filling in the post code into the url for a mapping application, or keywords for a search site<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"540\" src=\"https:\/\/alandix.com\/labs\/files\/2026\/11\/fig-2-2-onCue-1024x540.png\" alt=\"\" class=\"wp-image-561\" srcset=\"https:\/\/alandix.com\/labs\/files\/2026\/11\/fig-2-2-onCue-1024x540.png 1024w, https:\/\/alandix.com\/labs\/files\/2026\/11\/fig-2-2-onCue-300x158.png 300w, https:\/\/alandix.com\/labs\/files\/2026\/11\/fig-2-2-onCue-768x405.png 768w, https:\/\/alandix.com\/labs\/files\/2026\/11\/fig-2-2-onCue-1536x809.png 1536w, https:\/\/alandix.com\/labs\/files\/2026\/11\/fig-2-2-onCue-2048x1079.png 2048w, https:\/\/alandix.com\/labs\/files\/2026\/11\/fig-2-2-onCue-500x263.png 500w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p>At the time the web was still new for many people, so even being able to point out relevant sites was powerful, but onCue also aimed to make this as direct and easy as possible.  Now-a-days the &#8216;intelligence&#8217; seems simple, but was surprisingly powerful.<\/p>\n\n\n\n<p>onCue no longer runs, but the web-based <a href=\"https:\/\/aqtive.net\/demo\/onCue\/\" target=\"_blank\" rel=\"noreferrer noopener\">onCue (re-creation)<\/a> allows you to to get a feel for how it used to work.  The recognition engine has been rewritten in Javascript based on the Snip!t <a href=\"https:\/\/www.snipit.org\/tellmeabout\/\" target=\"_blank\" rel=\"noreferrer noopener\">TellMeAbout<\/a> service, which was itself based on onCue.  The interface tries to be as similar as possible to the original onCue, except, of course, floating at the side of a web page rather than the screen. <\/p>\n\n\n\n<p>As well as being a commercial product, onCue embodied several important interface and software design principles.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Appropriate intelligence<\/h2>\n\n\n\n<p>The term &#8216;appropriate intelligence&#8217; was coined in order to describe the design philosophy of onCue and has become a critical part of my own design vocabulary for AI, including in my books. At the core this is about designing focused on the entire socio-technical system not just the AI.  A critical consideration is the fact that AI is never perfect, neither in terms of how accurate or correct it is, nor in terms of whether it is relevant to the user, or in other words &#8216;<a href=\"https:\/\/www.alandix.com\/academic\/papers\/CHItaly2023-to-err-is-ai\/\" target=\"_blank\" rel=\"noreferrer noopener\">to err is AI<\/a>&#8216;.  This was captured in three rules for appropriate intelligence:<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>be useful when it is right <\/li>\n\n\n\n<li>be right often enough to be useful<\/li>\n\n\n\n<li>not cause problems when it is wrong<\/li>\n<\/ol>\n\n\n\n<p>The first two are what you expect for an intelligent system.  However, what really matters is the third.  Powerful AI can fail spectacularly when it gets this wrong (the sorry tale of Microsoft Clippy), but even simple AI can be useful if it is surrounded in an interaction that allows it to fail gracefully.  onCue endeavoured to do this.  <\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><span class=\"highlight\"><mark style=\"background-color:rgba(0, 0, 0, 0)\" class=\"has-inline-color has-blue-color\"><em>suggest not act<\/em>  <\/mark><\/span>\u2013  onCue only made suggestions, the user always got the final say.  If the system misinterpreted the clipboard content, or made suggestions that were not useful, the user could simply ignore them, rather than having to recover from unintended actions.<\/li>\n\n\n\n<li><em><mark style=\"background-color:rgba(0, 0, 0, 0)\" class=\"has-inline-color has-blue-color\">calm desig<\/mark>n<\/em>  \u2013  onCue was positioned at the edge of the screen and when it changed faded icons in.out over a period of about one second.  Peripheral vision is very sensitive to sudden changes, indeed one of its key functions is to attract your attention.  The slow fade meant the right icons for the current clipboard were always there when you looked but did not grab your attention when you were trying to do something else.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">Flexible architecture<\/h2>\n\n\n\n<p>onCue was built around a bespoke blackboard-style architecture (called aQtiveSpace) with small agents called Qbits (see <a href=\"https:\/\/www.aqtive.net\/community\/developers\/developers-pack\/onCue-hiw\/\" target=\"_blank\" rel=\"noreferrer noopener\">onCue: how it works!<\/a> for full details). and everythin was a Qbit, but its <\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img decoding=\"async\" src=\"https:\/\/alandix.com\/labs\/files\/2026\/11\/oncue.jpg\" alt=\"\" class=\"wp-image-568\"\/><\/figure>\n\n\n\n<p>Qbits were activated when appropriate data appeared within the onCueworkspace and could add items to the workspace.  This started with clipboard that would put a plain text item into the workspace.  Recognisers would pick this up and see if matched their rules, and if so put their own data back into the workspace.  Sometimes further recognisers would react to these, for example, the &#8216;words recogniser&#8217; would look for words in the text and if there were any deposit a &#8216;words&#8217; object including a bag of words, the &#8216;single word recogniser&#8217; would then be activated by the presence of a &#8216;words&#8217; item in the workspaces and if this contained just one word, it would deposit a &#8216;single word&#8217; data item in the space.  This chaining meant that more complex recognisers could be built up from simpler ones.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"667\" src=\"https:\/\/alandix.com\/labs\/files\/2026\/11\/Image38-1024x667.gif\" alt=\"\" class=\"wp-image-571\" srcset=\"https:\/\/alandix.com\/labs\/files\/2026\/11\/Image38-1024x667.gif 1024w, https:\/\/alandix.com\/labs\/files\/2026\/11\/Image38-300x195.gif 300w, https:\/\/alandix.com\/labs\/files\/2026\/11\/Image38-768x500.gif 768w, https:\/\/alandix.com\/labs\/files\/2026\/11\/Image38-461x300.gif 461w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p>Service Qbits reacted to particular data in the onCue workspace and create actions that would be fed into the Qbit that managed the onCue user interface.<\/p>\n\n\n\n<p>Other data detector systems were being developed at the same time, including <a href=\"https:\/\/developer.apple.com\/documentation\/datadetection\" target=\"_blank\" rel=\"noreferrer noopener\">Apple Data Detectors<\/a>.  However, most very closely coupled the underlying application (such as email), the detection through pattern matching and the final sue of the detected data (URL to oink or email address to launching email programme).  The onCue framework decoupled the underlying application being detected (by use of system-wide copy), the pattern matching (in recogniser Qbits), and the actions provided (through the service Qbits).<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Extensibility<\/h2>\n\n\n\n<p>This decoupling made it very easy to extend the system with new recognisers and services.  In particular recognisers were often more complex to write involving regular expressions (e.g. email) or custom code (e.g. tables).  In contrast, many services were simply placing the relevant components of a data value into a url.  So there were typically many services created for each recognised datatype.<\/p>\n\n\n\n<p>This also allowed relatively straightforward end-user extension.  Users could author simple Qbits using a wizard or by editing a simple XML descriptor file.  More complex recognisers could require custom Java code, but once this had been written multiple services could be added on top.<\/p>\n\n\n\n<p>This extensibility also meat that t was possible to create customised versions of onCue for commercial purposes, for example recognising order codes and customer codes and linkung these to relevant company intranet services. <\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"410\" height=\"330\" src=\"https:\/\/alandix.com\/labs\/files\/2026\/11\/oncue-order-screen.gif\" alt=\"\" class=\"wp-image-567\"\/><\/figure>\n\n\n\n<p>In many ways this is similar to integrated corporate systems such as SAP, but where the linking could cross heterogeneous subsystems &#8230; still a massive problem especially in the health service.  In later work I&#8217;ve used the term &#8216;<a href=\"https:\/\/alandix.com\/glossary\/aibook\/external%20semantics\" target=\"_blank\" rel=\"noreferrer noopener\">external semantics<\/a>&#8216; to techniques that establish semantic interpretations and connections between digital applications that have not been built into the original systems.  As well as data detectors, this include screen\/web scarers and now LLM-based agents.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The lattice of value<\/h2>\n\n\n\n<p>This was early in the days of internet application: the term &#8216;viral marketing&#8217; (in its current sense) only emerged during the latter part of the dot com period and marketing experts still talked in terms of conventional routes to reach customers such as magazines or advertising.  We recognised early on that products that market themselves within and between customer groups were a key way for small companies to gain entry into large markets with gargantuan advertising budgets!  Early product strategy used the term <a href=\"https:\/\/magisoft.co.uk\/alan\/ebulletin\/lattice-of-value\/lattice-of-value.html\" target=\"_blank\" rel=\"noreferrer noopener\">lattice of value<\/a> (and later <a href=\"https:\/\/magisoft.co.uk\/alan\/ebulletin\/market-ecology\/marketplace-1999.html\" target=\"_blank\" rel=\"noreferrer noopener\">market ecology<\/a>) to help us think abut these issues, and in particular to leverage website and application developers.<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/magisoft.co.uk\/alan\/ebulletin\/lattice-of-value\/images\/lattice.gif\" alt=\"\"\/><\/figure>\n\n\n\n<p>The extensibility itself allowed developers to create custom versions of onCue.  In addition we made it easy for developers to distribute custom versions of onCue with their own Qbits.  This included ensuring that the onCue licence was carefully worded to give developers confidence that they could distribute their versions without later finding that they or their customers would pay a fee, but also mean that in so doing we did not undermine aQtive&#8217;s ability to profit form onCue&#8217;s success.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">A sad end and a life beyond<\/h2>\n\n\n\n<p>aQtive&#8217;s original venture funding was only for the initial start-up phase and assumed a larger second round of funding would be required to come to market fully.  Around Christmas 1999 the company had numerous investments offers, but it was too soon. Of course in March 2000 the dot com bubble burst and the tech stock markets crashed following the floating of lastminute,com. Despite continued interest from investment managers, they ultimately came to a &#8220;we can&#8217;t invest in tech&#8221; brick wall within their organisations; so aQtive ended and with it onCue. As a last ghostly echo, for years afterwards, during conference presentations, one would occasionally see the little onCue icon floating on the screen.<\/p>\n\n\n\n<p>However, this was not the end.  Some years later some of the data detector ideas from onCue were built into the TellMeAbout service for the experimental web capture tool Snip!t, and a Javascript version of this was used to create the current onCue demo.<\/p>\n\n\n\n<p>I hope <em>sometime<\/em> to spend a bit more time to make the onCue re-creation easily extensible like the original (most of the recognisers and services are already defined using JSON specifications). If you would like to experiment with onCue yourself, give me a shout and I might accelerate that &#8216;sometime&#8217;!<\/p>\n\n\n\n<p><\/p>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>an intelligent internet interface from the dot com years onCue is referenced in many papers and also used as a case study in Artificial Intelligence \u2013 Humans at the Heart of Algorithms (2ed, Chapman and Hall, 2025) and AI for &hellip; <a href=\"https:\/\/alandix.com\/labs\/oncue\/\">Continue reading <span class=\"meta-nav\">&rarr;<\/span><\/a><\/p>\n","protected":false},"author":2,"featured_media":573,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_themeisle_gutenberg_block_has_review":false,"footnotes":""},"class_list":["post-554","page","type-page","status-publish","has-post-thumbnail","hentry"],"_links":{"self":[{"href":"https:\/\/alandix.com\/labs\/wp-json\/wp\/v2\/pages\/554","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/alandix.com\/labs\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/alandix.com\/labs\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/alandix.com\/labs\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/alandix.com\/labs\/wp-json\/wp\/v2\/comments?post=554"}],"version-history":[{"count":10,"href":"https:\/\/alandix.com\/labs\/wp-json\/wp\/v2\/pages\/554\/revisions"}],"predecessor-version":[{"id":583,"href":"https:\/\/alandix.com\/labs\/wp-json\/wp\/v2\/pages\/554\/revisions\/583"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/alandix.com\/labs\/wp-json\/wp\/v2\/media\/573"}],"wp:attachment":[{"href":"https:\/\/alandix.com\/labs\/wp-json\/wp\/v2\/media?parent=554"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}