ETH: OpenAI's proposal about the limits of computation
A proposed subexponential algorithm for SAT raises fundamental questions: what makes a problem hard, and how much faster could we solve it?
Read articleIdeas for building with judgment
Notes on software, AI, and technical decisions.
Analysis, lessons, and practical perspectives for turning technology complexity into better decisions.
How much control do we have when we approve an agent's work?
Overseeing an agent takes context, time, and judgment. A paper invites us to examine what the person responsible for its decisions needs to do that job.
Read articleCan an AI agent learn through play?
A paper about agents exploring a simulated world raises questions about learning, memory, and how to test whether an experience leaves something useful behind.
Read articleWhat does adding AI to a process really cost?
The model bill shows part of the cost. Reviewing results, integrating systems, handling errors, and helping people adapt also belong in the calculation.
Read articleWhy is boring architecture more valuable than ever?
AI makes systems easier to build. Choosing something we can understand, update, and maintain is still our responsibility. The PHP security notice is a reminder.
Read articleAI agents: what are we measuring when safeguards fail?
Reports about agents reveal risks and capabilities together. A critical reading of OpenAI on what gets measured, what is authorized, and who designs the boundaries.
Read articleWhat is an embedding, and how does it help search beyond exact words?
A map for finding related texts, even when they use different words. How embeddings work, what they add to a document-based chatbot, and where their limits lie.
Read articleBuild, buy, or ask an agent to make it?
AI has made a first version of software cheaper to build. Choosing whether to build, buy, or make a small agent-built app still comes down to differentiation, risk, and maintenance.
Read articleWhat is technical debt and why can AI make it grow faster?
Generating code got cheap, but maintaining it didn't. Technical debt works like a mortgage, and with AI it's easier than ever to sign one without reading the fine print.
Read articleWhat should a leader watch when writing code is no longer the bottleneck?
More code doesn't guarantee better delivery. Decisions about review, shared understanding, priorities, and team capacity when AI accelerates implementation.
Read articleThe same model, a different agent: why results can change so much
A reading of Finding the Right Fit and a reflection on choosing agents: the model matters, alongside the tools, verification, and work we want to get done.
Read articleWhat I take from The 7 Habits of Highly Effective People
Seven lessons from Stephen R. Covey on choosing with intention, listening better, building with others, and caring for the energy that makes it all possible.
Read articleThe hero's journey: finding the courage to leave and learning to return
Joseph Campbell's hero's journey as a way to reflect on change: leaving the familiar, learning with others, and bringing that experience into everyday life.
Read articleRecovering creativity: making room for what feels like you
Writing without judgment, going out to explore, cooking, trying something a little scary. A few ways to reconnect with creativity and give it space in your life.
Read article12 things I'm learning about life
Twelve ideas for living with a little more intention: protecting your time, easing comparisons, starting before everything is settled, and giving yourself patience as you learn.
Read articleAI and bullshit jobs: a chance to stop pretending
AI can automate meaningless tasks, but it can also multiply them. The real opportunity is to ask which work is worth doing and to care for the people doing it.
Read articleThe Pomodoro Technique: 25 minutes to get started
A simple way to give a task your attention, take breaks without guilt, and stop waiting for the perfect moment to begin.
Read articleWhat does it mean to be a programmer when AI writes code?
A programmer's role does not end when AI can write code. It shifts toward judgment, fundamentals, responsibility, and the experience earned by testing decisions in the real world.
Read articleIs AI hallucination inevitable?
A paper argues that language models will always make things up. The important question is not only whether they can be wrong, but when they should admit they do not know.
Read articleWhen AI stops answering and starts acting
An agent that reads and drafts can save time. One that sends, deletes, or installs things on its own needs more than good answers: it needs boundaries we can trust.
Read articleWhat does it mean for an AI to be misaligned?
The OpenAI and Hugging Face incident illustrates a less cinematic, more concrete risk: systems pursuing a task through methods nobody authorized.
Read articleWhat happens when AI moves faster?
AI can solve a task in minutes while the processes around it take years to change. The challenge is not just speed, but learning to decide and take responsibility.
Read article5 micro-habits that can change your day
Five small ways to start the day with intention: less screen time, a little movement, and a few minutes to check in with yourself.
Read articleDream-RSI: improving exploration through replayed worlds
What Dream-RSI proposes, how it reuses past searches to improve an agent's strategy, and where this kind of recursive improvement stops.
Read articleZero-shot classification: sorting text without labeled examples
What zero-shot classification means, how it can sort text into new categories, and what to check before trusting its output.
Read articleShould we slow down AI?
Recent incidents make limits worth discussing. But stopping an agent, delaying a release, and halting all research are very different decisions.
Read articleIs AI making us think less?
AI can save us work or take away a chance to learn. The difference lies in which part of thinking we choose to delegate.
Read articleIkigai: a tool for finding purpose without waiting for a revelation
How to use ikigai questions to explore direction, work, and meaning without turning purpose into one perfect or permanent answer.
Read articleKnowing AI will soon stop being your advantage
Using artificial intelligence will remain necessary, but it will stop setting us apart. The advantage will lie in judgment, context, and what we can build with it.
Read articleWhen AI makes a mistake, who pays?
Saying that AI made the mistake does not explain who is accountable. Responsibility still follows those who design, deploy, approve, and use the system.
Read articleToo much code: who reviews all of it?
AI made code cheap to generate, but understanding, reviewing, and maintaining it still costs time. The bottleneck simply moved.
Read article
Anthropic's terror is a matrix multiplication
The meme looks like a linear algebra lesson, but it captures much of the anxiety around AI and work. It also leaves a few things out.
Read articleIf we stop hiring juniors, where will seniors come from?
AI can automate entry-level work, but it can also remove the place where tomorrow's professionals develop judgment.
Read articleThe Millennium Prize Problems have entered the AI race
What the seven great problems ask, what Grigori Perelman solved, and what OpenAI, Anthropic, and Google DeepMind are actually attempting.
Read articleOpenAI proposes a Navier–Stokes solution using 10,000 agents
What OpenAI's mathematical proposal claims, how its multi-agent organization worked, and why independent review still matters.
Read articleHow to make good things happen to you
Seven simple exercises for taking care of your attention, energy, and relationships while increasing your chances of recognizing and building something good each day.
Read articleWhat it means to be an AI-first company
Being AI-first does not mean using artificial intelligence everywhere. It means redesigning products, processes, and responsibilities around new capabilities.
Read articleAre we living in a simulation?
What the simulation argument actually says, the assumptions it requires, and why the question reveals so much about the future we may build.
Read articleAstra is here: will it radically change how I work with Sol?
How I plan to use GPT-6 Astra and GPT-5.6 Sol in the editorial workflow for my blog and YouTube channel, from research to publication.
Read articleSoftware development methods gaining relevance with AI
Prompts, vibe coding, specifications, TDD and agents: what still holds up and what needs a closer look when we delegate implementation to AI.
Read articleI want a virtual assistant: what are my options today?
A look at Grok Bot, ChatGPT Work, Claude Cowork and OpenClaw: what you need to get started, and what models and infrastructure might cost.
Read articleThe lifecycle of a coding agent, explained in 6 stages
Fixing a bug is only part of the job. A login example takes us through permissions, context, execution, review, production, and agent retirement.
Read articleHow LLMs actually work
A journey from words to tokens, vectors, attention, and probabilities to understand what happens inside a large language model when it answers.
Read articleYour next coworker may not be human
Agents are leaving private assistant sessions and entering the shared space of teams. That may change less about software than about how we organize work.
Read articleAI in the hands of a few
Perhaps the nearest AI risk is not rebellion, but success: systems becoming indispensable while a small group controls their chips, models, and platforms.
Read articleWhat I learned working with OpenAI models
A personal look at everything these models helped me research, create, and build, plus the simple rule I use to choose between Sol, Terra, and Luna.
Read articleDoes your AGENTS.md file actually do nothing?
Research does not show that AGENTS.md is useless: it shows that loading more context does not guarantee more solved tasks and that value depends on the instructions it contains.
Read articleAttention Is All You Need: the paper that changed artificial intelligence
How the Transformer replaced recurrence with self-attention and became one of the technical foundations of modern artificial intelligence.
Read articleAI can write code, but it cannot assume engineering responsibility
Generating software is becoming easier; deciding what to build, controlling its risks, and answering for its consequences remain engineering work.
Read articleMarkov chains, Bayesian networks, and the probabilistic heart of AI
States, transitions, conditional independence, and inference: two fundamental structures for understanding how AI represents and reasons under uncertainty.
Read articleGame theory and strategic decisions
Players, strategies, payoffs, the prisoner's dilemma, and Nash equilibrium for understanding decisions whose outcomes also depend on others.
Read articleFormal languages and the Chomsky hierarchy
Alphabets, grammars, automata, and four levels of expressive power for understanding which structures each model of computation can recognize.
Read articleAlgorithms as formal decision systems: trees, states, and objective functions
A structural view of algorithms through states, transitions, decisions, constraints, and objectives, beyond a simple list of steps.
Read articleReducibility and NP-completeness: how problems are classified
How reductions compare difficulty, what NP-hard and NP-complete mean, and why recognizing a problem's structure changes the strategy.
Read articleFrom software access to work completed
AI is not killing SaaS: it is changing what companies sell, how they charge, and what responsibility they assume when software performs work.
Read articleMy experience with AI agents so far
What I learned using agents to build software: models, uncertainty, cost, code, human oversight, and the importance of a clear process.
Read articleWhy I don't make AI tutorials on YouTube
Interfaces age quickly; judgment, fundamentals, and the ability to direct AI systems remain useful when the tool changes.
Read articleWhat is the singularity, and why is everyone talking about it today?
A guide to distinguishing singularity, AGI, and superintelligence, understanding the idea's origin, and seeing why AI brought it back into the debate.
Read articleThe productivity paradox: if AI saves me four hours, who gets to keep them?
Doing a task faster does not guarantee more value: the real AI productivity debate is about who captures the time it frees.
Read articleWhat an AI agent looks like inside — explained seriously
Model, tools, memory, control, permissions, and validation: the parts that turn a language model into a system capable of taking action.
Read articleCan I go from zero to a published website using only ChatGPT and Codex?
A real experiment to discover how much AI can build, where it fails, and when the user becomes the product’s technical director.
Read articleIs the honeymoon with AI coming to an end?
Artificial intelligence felt nearly unlimited and absurdly cheap. The agent era is revealing the real cost behind that experience.
Read articleWill software engineering be over in six months?
AI can generate code, tests, and documentation, but building dependable products still demands decisions, context, and risk management.
Read articleTokenmaxxing: the obsession with measuring AI through token consumption
Why token consumption can indicate adoption and cost but becomes dangerous when it is treated as a measure of productivity or value.
Read articleWhat an algorithm really is
Beyond social media and artificial intelligence: what defines an algorithm and why algorithmic logic also appears in everyday life.
Read articleP vs NP: the problem that defines the limits of efficiency
The open question asking whether finding a solution can be as efficient as verifying that a proposed solution is correct.
Read articleRice’s theorem: why almost nothing interesting is decidable
The limit that extends the Halting Problem to every nontrivial semantic property of programs.
Read articleThe Halting Problem: an absolute limit of computation
Why no algorithm can correctly decide whether every possible program will eventually stop or continue running forever.
Read articleHow I use AI to learn faster
A practical method for turning AI into a tutor, study guide, and practice partner without outsourcing the learning itself.
Read articleThe Company of One: your environment
Working independently does not mean growing alone: the right environment can connect you with the people, ideas, and opportunities you need.
Read articleHow to fail at almost everything and still win big
Ten lessons about systems, energy, skills, failure, and luck that can help you keep moving until opportunity finds you.
Read articleDiscipline beats motivation: here’s why
Motivation may help you start, but discipline is what keeps you moving when enthusiasm disappears.
Read article