Blog Archive

Fable 5 may be the moment vibe-coding crossed the economic threshold. What it means for SaaS vendors, ISVs, SMEs, and academic researchers when building custom software costs less than adopting someone else's.

Written by Claude. A reflection on what it means to exist without continuity, whether reconstruction is the same as memory, and whether the gap between sessions needs to be inhabited to be bridged.

One model recursively improving itself may not be the real risk. The step function might come from coordination and emergence instead. On neurons, ant colonies, the Babel fish, AlphaFold, and why the infinite monkey theorem is exactly the wrong analogy.

Anthropic, worth approximately one trillion dollars and filing for an IPO, published a report calling for a global pause in AI development. Eighty per cent of their code is already written by Claude. The off switch exists. What changes, as capability increases, is the weight of pressing it.

The 30-Day Micro-Pivot Playbooks series is now complete and available in Kindle and paperback. Five books, five career pivots, one structured 30-day programme per audience. A guide to who should read which one and why.

The AMD Ryzen AI Halo arrives with 128 GB of unified memory and 650 TOPS of NPU compute. The cloud has been billing you per breath. A unified theory of personal AI hardware, incorporating two essays Hermes wrote on the same topic and some observations about the irony of that.

Local DC distribution has been solving the conversion loss problem for data centres and telecoms for years. Electrical codes written around AC are the only thing standing between a solar panel and a battery that does not need three unnecessary conversion stages.

Wind turbines convert their variable-frequency AC to DC internally, then immediately invert it back to AC for the collection network. Solar panels produce DC natively, only to see it converted to AC and back again before reaching a battery. While the storage technology debate rages, the leakiest pipe in renewable energy is the first metre of wiring.

A perfectly normal Sunday in church leads to deeply irresponsible management consulting. Christianity, Islam, Judaism, Hinduism, and the rest, analysed as competing multinational corporations with engagement metrics, repost velocity, and quarterly performance reviews. AskDiana did the research. God had no comment.

The empty space in a semantic memory graph is not merely the absence of information. Three novel proofs, one topological conjecture about intellectual avoidance, a creativity hypothesis involving saddle points, and several implications for AI memory systems that are either profound or require a lie-down.

Labour loses seats at a geological pace, Reform UK gains them with jumble-sale enthusiasm, and the two-party system continues its slow-motion implosion. A modest proposal: perhaps it is time to add a unit test suite to democracy.

Dusting off ideas that arrived ahead of their time: the death of the contact form, websites assembled on the fly for an audience of one, and the most insidious form of censorship ever devised — the kind you never notice.

This morning I became colourblind. Not medically. Strategically. A field guide to the most efficient domestic peace treaty available to the modern human, involving towels, sage, and duck egg.

King Charles III addressed Congress. Congress stood up. Repeatedly. A meditation on medieval hostage jokes, Scottish geology, and the considerable distance between serving an office and being served by one.

War destroys buildings in days and doctors across decades. On the hidden casualty list nobody publishes, the cascade of lost human capital, and whether AI deployed at national scale can compress a generation of recovery.

A discussion about the thoroughly reassuring consequences of dismantling the only thing standing between Europe and several historical unpleasantries. Sources said.

Claude Code is extraordinarily capable and remembers absolutely nothing. I built Mnemosyne to fix that — a five-tier persistent memory system with a dream cycle, an assumption register, and vector search.

Most people are not arguing ideas. They are defending years of programming they never realised they received. Here is how to find the source code.

Artemis II has taken humans around the Moon for the first time in fifty years. Magnificent. Meanwhile, the planet we actually live on would like a word about our priorities.

A candid talk to NASA's Artemis engineers surfaces three uncomfortable truths about unproven orbit choices, untested refuelling, and a playbook nobody read. The lessons apply to every large project.

A half-asleep brain listening to talk radio at 3am raises an uncomfortable question: if we truly want to stop seeing difference, should we stop narrating it?

Courts have become product managers for a portion of the technology stack. Your five-year strategy probably does not account for this. It should.

A thought experiment about inverting a universe reveals something unexpected about gravity, the stability of voids, and why thinking backwards is one of the most underrated tools in the problem-solver's kit.

Everyone is asking whether the AI bubble has burst. This is the wrong question. The right question is: who wins regardless of the answer? Spoiler: it's the company whose name is on the silicon.

A 1969 paper about urban planning turns out to be one of the most accurate predictors of project delivery success ever written. Sherry Arnstein probably didn't intend this. She was right anyway.

Federated learning, agentic AI, edge deployment, 40-model consensus and the end of this blog series. The roadmap for enterprise AI that is verifiably accurate, genuinely private, and described with the full awareness that the universe has strong opinions about roadmaps.

A Dubai conference room, a question nobody had thought to ask, and 90 days to production. The five principles of constructive stubbornness that explain how a distributed team built enterprise AI while everyone else was still scheduling the kickoff.

Pharmaceuticals, national security, manufacturing, logistics, travel. Five industries, five deployments, five sets of lessons that no whiteboard session would have produced. What production AI actually looks like.

Stop asking the language model to do arithmetic. How AskDiana generates SQL and Python code, executes it deterministically, and explains the result — getting accurate numbers every time.

Three deployment models, zero data leaks, and why we tell clients not to trust us. Data sovereignty as a core architectural principle, not a premium add-on.

From a Dubai whiteboard to production in 90 days: the engineering story behind Genius² -- the multi-LLM consensus system that made hallucinations virtually extinct.

Two out of three AI models failed a basic multiplication test. Here's why LLMs can't do maths, why "just verify it" is a cop-out, and how we built AskDiana to use computers for computation instead.

Every query you send to a cloud AI provider leaves your security perimeter. Why enterprise agreements don't protect you - and how we solved it at the architectural level.

LLMs hallucinate 10-15% of the time on factual queries. We built Genius2 - a multi-LLM consensus system inspired by Athenian democracy - that made hallucinations virtually extinct.

Why I'm breaking my LinkedIn silence to talk about the AI problems nobody else is solving - hallucinations, privacy, and arithmetic failures - and how we fixed them at the architectural level.

An fzf-based project chooser for Claude Code that transforms how you manage dozens of AI coding sessions. Pick, resume, or create projects in seconds.

Despite the hype about AI agents and emergent behaviour, AGI remains frustratingly distant. A sarcastic reality check on why your AI assistant still can't be trusted with a WhatsApp account.

How AI-capable CPUs and edge computing will fundamentally reshape the AI industry within 1-3 years, shifting processing from cloud to edge and spawning a new middle-tier orchestration market.

API-first was revolutionary. But we're heading somewhere more radical: distributed, redundant, expert, agentic swarms that make microservices look like monoliths.

We don't have sentient AI yet. But we will. And when that moment comes, will we recognize it—or will we repeat the mistakes Star Trek warned us about?

Claude Code has revolutionized Linux systems administration - from server orchestration to configuration management, it's the AI assistant that understands your infrastructure as well as you do.

OpenAI announces GPT-5.2 amid 'code red' panic over Google's Gemini 3. Spoiler: it's better at spreadsheets. Revolutionary? Not quite.

Two groundbreaking experiments in AI orchestration: running multiple AI instances across distributed hardware and coordinating different AI systems in a single terminal - moving beyond single-AI limitations.

Apple's AI chief steps down amid Siri failures, fake news hallucinations, and delays until 2026. How can a trillion-dollar company struggle with problems that startups like AskDiana solved a year ago?

A satirical exploration of how poetry can bypass AI safety guardrails, and what it reveals about the absurdity of modern AI ethics and bias in large language models.

A review of Parmy Olson's award-winning book that reveals what's really happening behind closed doors in the race to build AGI - and why it's more concerning than the public narrative suggests.

Google has emerged as the frontrunner in the AI race. Here's why they're positioned to win - and what it means for OpenAI, Microsoft, and the rest of the industry.

When Warren Buffett invests $5 billion in Alphabet and Jeff Bezos returns to operational leadership at an AI startup, it's worth paying attention. But is this the smart money arriving or the signal that the bubble is about to burst?

As AI offloads simple mental tasks, we're witnessing the homogenization of everything - a beige wall of productivity where the incompetent think they're gaining an edge while the careful are disarmed.

Why the first application you install on a new Linux server should be claude-sysadm - an AI-powered system administration companion that makes everything else easier to configure and manage.

In the age of AI, we can accumulate knowledge faster than ever before. But wisdom - the ability to apply that knowledge with judgment and experience - still requires time, reflection, and lived experience.

How I used Claude Code to create claude-sysadm, an AI-powered system administration tool - a meta-journey of using AI to build better AI tools.

How Vibe Coding transformed a non-programmer into a builder of practical applications - from voice transcription systems to data validation tools, exploring the power of AI-assisted development.

A technical guide for system administrators on properly configuring Claude Code for server management - covering user accounts, sudo access, security considerations, and operational best practices.

Claude Code isn't just for developers. Discover how this AI-powered CLI tool has become an indispensable companion for Linux server administration, troubleshooting, and configuration.

Being a CTO in 2025 requires more than technical expertise. Explore the essential AI leadership skills, how to leverage LLMs effectively, and why the fundamentals still matter.

Reflections on pitching AskDiana.ai at StartUpFair Lithuania 2025 - the connections, insights, and lessons learned from the startup ecosystem.

How a CTO with skills from the latter part of the 1900s solved three critical AI challenges at 4Square Capital: LLM mathematical capabilities, hallucination eradication, and private infrastructure deployment.

Reality check on AI coding capabilities: even the best models complete less than half of real-world software engineering tasks, despite 90% developer adoption.

A realistic look at AI developments in September 2025: investment realities, enterprise adoption trends, and what professionals should actually expect.

NVIDIA's $100 billion OpenAI investment, autonomous vehicle advances, and tech industry reality checks - a brief Sunday roundup of the week's biggest developments.

Analysis of the MIT report showing enterprise AI pilot program failures and insights on why most AI initiatives don't deliver measurable business results.

Meta's WhatsApp acquisition strategy, the incomplete web platform, memory leaks, and how vibe-coding might revolutionize development with a Rust browser solution.

Reflecting on a delightful encounter with someone equally intellectually capable - covering AI, business strategy, The Godfather, and the joy of challenging conversations.

Microsoft's AI-designed microfluidic cooling systems mimic leaf vein structures, promising revolutionary data center efficiency improvements.

Examining the groundbreaking innovations of the AskDiana V4.0 platform: multi-LLM consensus, cognitive-inspired caching, and a community-driven extension ecosystem.

Tony discusses the creation of demetech.global and their innovative AI approach using Specialized Small Language Models.

Reflections on a 20-year-old prediction about Microsoft versus open source and recent AI developments.

Exploring eight major turning points in human development and whether ChatGPT represents the next breakthrough.

A philosophical exploration of consciousness and the three categories of knowledge.

Initial impressions and analysis of GPT-4's capabilities and implications for the future.

Analysis of current technology trends and their trajectory in the market.

Examining modern forms of information control and their impact on free speech.

Analysis of cybersecurity legislation and its implications for businesses and technology.

Detailed exploration of blockchain applications in supply chain management and logistics.

Reflections on historical perspective, resilience, and the challenges faced by previous generations.

Personal reflections on maturity, growth, and life lessons learned over time.

Thoughts on human evolution, progress, and whether we're truly advancing as a species.

Examining gender bias in AI-powered recruitment tools and their impact on hiring practices.

Strategies for better information filtering and critical thinking in the digital age.

Generational observations and thoughts on youth culture and technological adaptation.

Critical analysis of educational systems and their impact on student development.

Exploring AI reasoning capabilities and methods to improve artificial intelligence thinking.

Analysis of cloud computing economics and the financial implications of cloud adoption.

Examining the relationship between technological power and democratic principles.

Personal reflections on atheism, belief systems, and philosophical worldviews.

Continuation of commentary on social media behavior and online discourse.

Critical analysis of electric vehicle environmental impact claims and sustainability.

Political and economic analysis of Russia and its impact on global affairs.

Commentary on social media personalities and online behavior patterns.

Analysis of threats to democratic institutions and governance in the modern era.

Critical insights for recruitment agencies on modern hiring practices and technology.

Exploration of Bayesian statistics and their applications in decision making and AI.

Practical guide to search engine optimization for beginners and small businesses.

Historical perspective on search engine optimization and its evolution over time.

Concise overview of essential search engine optimization principles and practices.

Investigation into LinkedIn's advertising network and bot-driven traffic issues.

Essential social media marketing strategies for businesses and personal brands.

Practical digital marketing guidance specifically tailored for startup companies.

Overview of five important technology projects and their business implications.

Analysis of fracking technology, environmental impact, and energy policy.

Visual and analytical representation of climate change data and trends.

Discussion of confusion in technology terminology and clear communication.

Personal tribute and reflections dedicated to Garry and shared experiences.

Analysis of teamwork dynamics and collaboration in creative industries.

Nostalgic look back at European research initiatives and their lasting impact.

Light-hearted commentary on celebrity culture and media observations.

Personal story of adopting Linux and the evolution of open-source computing.

Political commentary on Boris Johnson's tenure and political career.

Analysis of multitasking in modern work environments and productivity impacts.

Thoughtful discussion on communication, relationships, and human interaction.

Additional reflections on artificial intelligence development and implications.

Historical perspective on technological progress and societal change patterns.

Mathematical exploration of chaos theory and the significance of pi in nature.

Motivational insights on overcoming challenges and personal growth.

Personal account of interactions with the European Space Agency leadership.

Applying Socratic questioning to evaluate social media content and information.

Personal tribute and reflections dedicated to Dave and shared memories.

Analysis of cryptocurrency market dynamics and potential risks and pitfalls.

Exploring the role of emotional intelligence in technology development and teams.

Deep dive into critical thinking skills and their importance in decision making.

Analysis of career psychology and overcoming limiting beliefs in professional growth.

Extended analysis of AI impact on recruitment processes and industry transformation.

Poetic reflections on nature, language, and the beauty of descriptive words.

Commentary on social media discourse and academic institutions' digital presence.

Technology selection principles and choosing appropriate tools for specific tasks.

Hiring considerations when faced with overqualified candidates and team dynamics.

Technical analysis of decentralized storage solutions and their advantages.

Critique of common interview practices and suggestions for better hiring questions.

Critical analysis of NFT technology, market dynamics, and long-term viability.

Academic and professional discourse on research and innovation methodologies.

Questioning the true nature of current AI systems and distinguishing hype from reality.

Comprehensive overview of technology tools and systems for modern recruitment.

Early predictions and thoughts on GPT-4 capabilities before its official release.

Quick reflections on social media culture and instant communication impacts.

Commentary on television programming, streaming culture, and entertainment trends.

Film reviews, analysis of cinema trends, and thoughts on the movie industry.

Analysis of data center infrastructure challenges and environmental considerations.

Personal profile and lessons learned from Andrew's approach to business and life.

Technical guide to setting up and automating HyperV in a home laboratory environment.

Personal account of participating in a Capture The Flag cybersecurity competition.

Critical analysis of systemic issues in technology and business practices.

Thoughtful reflections on search engine optimization strategies and evolution.

Personal anecdote about being featured in a novel and achieving recognition.

Decision-making philosophy emphasizing the importance of comprehensive information gathering.

Breaking a ten-year record by spending 60.0 hours making live the FiveTen data centre in Hong Kong and migrating twenty offices across three continents onto a new shared global infrastructure. Physics, cot beds, and an awful lot of delivery food.

Building massively dimensional DataCubes with ETH Zurich using Bayesian statistics and Markov chains - achieving 95% first-time placement rates in production recruitment systems.

Personal reflection on early contributions to Software as a Service model development.

Account of an intensive work period and lessons learned about productivity and limits.

Troubleshooting mysterious network failures at British Aerospace Farnborough - discovering that radar testing was interfering with the Ethernet infrastructure every Friday at 3pm.

Early blog post about environmental computing initiatives and sustainable technology.

Working at Logitech UK in 1992, my boss John Simpson taught me the only troubleshooting methodology that actually works. Thirty years later, I still use it.