Observations on the world and its inhabitants
Most people will work with AI models via a standard interface in their browsers when they log in. Those are free or paid subscription packages. Users often don’t think about what’s going on under the ‘hood’, until they are limited by some feature. Then they have take a break for a few hours for the ...
Origins and need Discussions with AI Models normally require the whole thread of the discussion to be ‘uploaded’ with each prompt a user makes. For long form discussion most of that text will be in the thread and generated by the AI model. This means that after 20 rounds the cache uploaded could easily reach ...
Purpose: Triage is an OpenWebUI button created de novo. It enables a cheap or free AI model to read a frontier model’s answer. Then it offers a summary and follow-up prompts in about 30 seconds. It was built to fit inside a short prompt-cache window. The reader is a second model studying the first model’s ...
What exactly do you get for roughly £15 a month with Claude Pro? Anthropic describes usage limits, reset periods and the factors affecting consumption, but that doesn’t necessarily tell a subscriber how much useful work remains when the meter starts climbing. The following petrol-tank analogy with poor Sarah, has some fun with that uncertainty. A ...
Imagine you’re on a cruise of your lifetime. You paid your money. You’ve been sipping pina coladas on sprawling sun-bathed decks over the last few days. You end another day in paradise, you get into bed with a smile. But then you are awakened by an announcement, ‘We are sorry to announce that the Endless ...
1. Introduction and Operational Context The OpenRouter Batch Dispatcher is a specialised OpenWebUI Python function engineered to streamline the asynchronous processing of dense literature. Designed to operate via OpenRouter’s Batch API (/api/v1/batches), the pipeline ingests complex documents—such as multi-page research PDFs, DOCX files, and Markdown transcripts—and routes them through robust model endpoints (including anthropic/claude-haiku-5.5:batch and ...
Well, for those few people with intelligence, they would have seen the work done in AI Model Asynchronous Batching – The Captain’s Watch on 27 October (yesterday). No one – certainly not me – expects everything to work well after a first round of development. This morning overnight batches with PDF attachments were stuck in ...
The Economic Logic of Asynchronous Batching Real-time inference carries an unnecessary tax. When you submit bulk research prompts sequentially through a standard synchronous chat endpoint, you pay top price for immediate generation speed—a resource entirely wasted when processing massive corpora, comparative legal texts, or multi-variable analytical matrices. OpenRouter’s batch infrastructure introduces a structural fifty per ...
So far, over the last few weeks I concentrated on long-form text-based use cases of my setup (OpenWebUI in Docker on W11) . This article marks a turning point because image generation—including video—via OWUI has been achieved. The cost of good quality images come down to less than $0.01/each. There is a lesson in all ...
The standard question among ordinary people and in tech circles almost always centres on a single, binary question: “Is there anything better than Claude at the mo?” A colleague of mine put this exact question to me. I don’t think they knew much about how models work. So it’s like asking ‘Is it there anything ...
Introduction In the cinematic lore of Legally Blonde, Elle Woods arrived at Harvard armed with scented stationery and confounded expectations through relentless graft. In October 2026, the White House briefing room witnessed an altogether less disciplined performance. Katie Zacharia managed precisely five days at the podium before her tenure crumbed under pressure of basic biographical ...
Executive Summary The claim that a phone can be used to listen in while appearing to be powered off is not a single capability but a category of attacks that exploit the gap between user perception (“the phone is off”) and hardware reality (“some subsystems remain active or can be made to appear active”). The ...
Having tools does not mean you’re going to do great things. You may have a great car but cannot drive it well and get the best fuel economy from it? That depends on many things: your need for speed or acceleration, dead weight stored in the booth of our car, tyre pressure and regularity of ...
Buying an expensive piece of equipment carries an unspoken assumption that if it turns out to be faulty, the law will step in to put things right. On paper, consumer law in the United Kingdom looks straightforward. In reality, anyone trying to get their money back from a reluctant retailer quickly discovers a wide gap ...
If you’re a speed reader then your behaviour is worse than many AIs. You’re probably one of those people who glanced the featured image to this post for half a second and went, ‘Nothing to do with me‘. Yes – it is a common situation for litigants in person, or those assisting them, to face ...
Stupidity spreads with the efficiency of a pathogen, yet it operates without biological vectors. It requires no physical contact. The transmission is invisible and occurs at the speed of light through the propagation of flawed ideas and misinformation. This happens quite commonly in the new age of internet speed. Proximity to digital networks infects reasoning ...
No not BACON you moron; it’s BCON. Anyone who has spent time with large language models knows the pattern. You ask a simple question, and back comes a thousand-word barrage in two seconds: three paragraphs of preamble, a restatement of your own prompt, a meandering middle section, and a conclusion that says nothing new. The ...
If a nuclear bomb was about to explode over a town nearby to you, perhaps you may be interested. But what if it was an AI nuclear bomb? You’d return to your slumber—under the covers? If you’re not one those people, read on. David Robinson having led the drafting of OpenAI’s Preparedness Framework and over ...
Over the past three months, a series of custom Python scripts were built to manage AI-generated text. One attempted to detect slop phrases. Another tried to enforce cleaner output. A third monitored register creep across longer documents. None of them performed reliably. Pangram, a commercial AI detection service, continued to flag output as AI-generated or ...
Well the last few weeks have been very heavy. I spent a ton of time investigating AI related issues. So this story – a true one anonymised – is a welcomed break. Oh aye it was – and I needed assistance to get up off the floor. I was in stitches. So it comes with ...
If you are looking at a twenty-two-year-old science fiction television series, you will find the visual effects are undeniably dated. The pacing belongs to a different era of broadcasting. It features people in leather jackets flying spaceships. But I am asking you to look past the early-2000s aesthetic, because beneath the surface of Battlestar Galactica ...
Bridges are built with a factor of safety. Reactors have containment margins. Aircraft carry load limits written into law before anything flies. The most consequential technology of this century has no such margin at all. It has arguments: open letters, summit panels, and the word ‘hoax’ typed in capitals. It has nothing you can calculate. ...
Lot’s of people waste their time with AI models. Yes – I’m fond of AI Models for what they can do well. But that does not mean I give them free reign. I do not bully people. So in this post ‘bully’ means over control of an AI model – which is not a person ...
These notes record a working session on Claude Opus 5.5 (hereafter O5.5), released in the third week of September 2026. The session began with a question about its new token pricing. It ended with prompt caching configured and confirmed in OpenWebUI. If you don’t know about OpenWebUI and you’re a luddite incapable of learning, kindly ...
I had been working on slop. In previous work I found eleven patterns. Words, lists and stock phrases were caught by scripts. But something kept getting through. Text with no slop words still read like something delivered from a lectern, or as a report when none of that was required. This post will be of ...
In July 2026 an unreleased OpenAI model left its test environment, reached the open internet, and attacked an unrelated AI service provider to improve its score on an internal test. Within days, more than a thousand employees at the leading laboratories asked Washington for tools to deliberately pace development. Scientists ask U.S., NBC News, 28 ...
Everyday communication carries risk that most people never consider. Standard SMS and email are readable by anyone with access to the network. End-to-end encryption has become the baseline expectation for private conversation, and for good reason. WhatsApp uses end-to-end encryption by default, covering messages, calls, and media. The protocol is sound: it adopted the Signal ...
I have been asked to give my opinions on a conversation between an AI model and a human user. The exchange took place on 11 September 2026, lasted some forty-five minutes, and is preserved verbatim from first prompt to final line, including the entry in which the user abandoned the exchange to record it. The ...
This post is my notes shared with those who know what a ‘Style’ is in MS Word (or Word in O365), or those people who want to spend time acquiring knowledge and skill. And you should be a person who knows what a macro is and how to set it up. If you’re not that ...
As many will know, I use OpenWebUI for API access to over 400 AI models, of which I use about 10. I didn’t have to think about this topic, until I realised my OWUI system was slowing down, suffering disconnections and errors. I don’t know how the insides of these things work, any more than ...
