Anticodeguy

<written by a human being>

The time has come when you no longer need to look for the proverbial computer master to fix your Windows. With the help of agents such as Claude Code and Codex, which work from the command line in the system, you can easily perform all necessary checks and corrections to the system that will provide proper maintenance, fix malfunctions, and restore normal operability. In chat mode it was also possible to do this, but you had to manually type commands and send the results of execution in the command line so that AI understood what was happening and suggested what to do next. Literally just a few months ago I did exactly that, but now I simply launch Claude Code from the command line, where it works independently: runs system health checkup and recovery procedures, checks the registry for junk entries, checks the system for unused clogging components, cleans temporary files. It even suggests deleting programs that you possibly don't even use anymore. Recommendation: everything will be simplified if you launch Claude Code from PowerShell as administrator. Then you won't need to manually confirm some actions that require admin access. The prompt is on the screenshot.

Your business plan is a lie – and that's okay Read more about The Three-Body Problem: Why Your Business Dreams Keep Crashing into Reality

<written by a human being>

Lately I've noticed that every day I'm looking for tasks for myself that I'll be able to complete with the help of artificial intelligence. And the first question I ask myself when I start a new task is whether I can solve it by passing it to Claude Code, the new Cowork feature, or just in chat. And oddly enough, tasks that I can perform with the help of AI agents seem even more interesting, because it really feels like magic. You give the agent as input a bunch of tables with data for several million records, explain to it how they need to be processed, changed, and assembled into a configuration at the output. And then you watch as it engages in programming in Python, tries to complete its task, finds errors, fixes them, and ultimately produces a ready result. You check and the task turns out to be completed. In 20 minutes you do with it what would have taken you several days with a programmer, with whom you would have needed to exchange technical specifications, corrections, quality assessments, and other joys of human intellectual labor. Who else feels the same way?

<written by a human being>

If I hadn't caught the erroneous direction during the production process, the AI would have calmly deployed the client's knowledge base with a different authorization system, and all of this would have had to be fixed after the fact. Naturally in development, the earlier you notice the incorrectness of a pattern, the easier it is to fix. At an early stage, it always costs much less blood. Therefore, besides the fact that a human must be present during development by artificial intelligence, they also must be knowledgeable in these matters and understand what's happening. This is exactly why the best users of AI agents when writing code are still developers who can write and read code, who understand architecture and know at what moment it's better to interrupt an overzealous agent that without warning can rush to execute the next stage with an incorrectly completed previous one. A good reminder that artificial intelligence is still a tool, and not yet a full replacement for a human. And in reliable and skilled hands it becomes a superpower. But in the hands of a person who doesn't account for their actions, naturally, it can cause significant harm.

Why every business blueprint you've bought has failed you Read more about The Three-Body Problem: Why Your Business Dreams Keep Crashing into Reality

<written by a human being>

The peculiarity of AI work is that it forgets certain things. A concrete example that I had recently in a project deploying a knowledge base for my client. Before starting the project, we developed the architecture with the AI agent and documented it. In particular, a specific user authentication model was defined through Authentic - a unified system that would allow both my agency employees and clients to authorize into different systems with different access levels, including the knowledge base and the product management tool. But when deploying the knowledge base itself, we encountered a couple of errors and went around them for some time to fix them. The next step was authorization, and the agent began making an absolutely different system that wasn't connected with the previously agreed architecture. I had to stop it and remind it about the agreed model. In such cases, as a rule, I start a new chat, a new session with the AI agent, because it's clear that the context window isn't sufficient at this moment. This is exactly the signal to reset it, which happens in a new chat.

<written by a human being>

Without a human as an observer of artificial intelligence when it writes code, we still can't do without one for now. Overall, the code itself it writes is normal, but very often it makes decisions that would be better to correct. And there are still many moments when during development the agent loses context and simply forgets about the decision that was made earlier. As strange as it may sound, human memory still works better in this regard than the contextual memory of an AI-agent. No matter how large it may be, it's still very limited for now.

It’s time for human watermarks on content. When we generate content using AI, they automatically place a watermark indicating that it was generated by Sora or OpenAI, or Midjourney. Each model has its own distinctive marks. When we generate content using people, we don’t place anything. But since soon - and quite possibly already now - more content will be generated specifically by artificial intelligence, I believe it makes sense to start placing such watermarks on content that was written and created by a human, because it now carries greater value. This post was written by a human.

The hard truth about doing what you love: nobody cares about your passion until it solves their problems Read more about The Ikigai Blueprint: Finding Work You Love That Pays You Well

<written by a human being>

How are no-code solutions feeling in the age of AI? Today there's nothing left but to continue manually editing and fixing systems implemented on the basis of no-code platforms. Some of them already allow partial configuration through artificial intelligence, but only if the platform itself embeds AI into its engine, which, for example, Bubble is doing. All of this, in my opinion, will lead to either new builders appearing that are AI-first, which should become the logical continuation of all these no-code solutions, or they themselves will begin to remake their architecture to primarily be tailored for working with AI agents. But this is in any case an order of magnitude more complex, because it's essentially changing the foundation under the house, which implies its complete disassembly down to the very base. Well, and the last option - they completely die from lack of necessity, since it's already become easier to write prompts in native language and get a ready product than to mess with endless workflow settings or craft responsive design in a drag-n-drop editor. What do you think will happen?

<written by a human being>

Here's what I've noticed about products used in work: those parts of the information system that can be edited using the command line, including server settings and everything that happens inside it, are, in principle, accessible to artificial intelligence. You just need to provide the appropriate access. But to edit an application on Bubble, no access will help. And even if you simply ask artificial intelligence about how to fix one issue or another, the answer in almost 100% of cases will be incorrect, because there's a huge number of nuances in the platform itself that are unknown to artificial intelligence, because they simply weren't documented in any way, and simply aren't described anywhere. Thus, the answer will simply be a hallucination. And secondly, since it's a builder, there's a lot of specificity there that relates specifically to the current project. From a builder you can assemble any Frankenstein's monster, and it's impossible to account for all the nuances that will emerge with one refinement or another from a simple technical specification or indication of an error.

You don't find your passion. You build it. Read more about The Ikigai Blueprint: Finding Work You Love That Pays You Well

<written by a human being>

Curiously, it seems that in the world of web development, the pendulum has swung from builders and no-code solutions back toward code. For the reason that AI agents can't fully work with no-code tools, since these are closed platforms, and they don't give access to their source code, don't allow editing everything through the command line. And simply, using text in one way or another, which is the key method of interaction with AI, doesn't allow fully creating solutions and products there with the help of artificial intelligence. But at the same time, it's now become even easier to write some script or application in code. Because from a developer's perspective, it's already easier to write a prompt in your own language than to dig through complex builder interfaces. You write a series of needed prompts, compose a technical specification that will be fully taken into account, and after several iterations with corrections, everything will be executed and implemented on a code base.

<written by a human being>

Who else is tired of endless monthly subscriptions and paid SaaS in the age of artificial intelligence? When I'm increasingly using free open source tools because I can simply deploy them for myself easily with the help of AI agents, or just write the software I need for work myself. I thought, why don't we return to the roots, to good old software that you pay for once and use for life. Yes, maybe it's not as interesting from a business point of view, but how cool is it from a user's perspective! You don't need to worry that you'll be charged every month and budget for it. No, you simply lay out a small amount of money once, which, as a rule, doesn't exceed the cost of a monthly subscription to modern services, but you need to pay it not every month, not every year, but simply once. And then you have a working application that may not even be connected to the internet at all. And there won't be a surprise in a month when you suddenly wake up and can't use its functions. This is exactly the path I decided to take. The application can be purchased by paying for it once. Pretty crazy for 2026, right?

You've been lied to about "following your passion" Read more about The Ikigai Blueprint: Finding Work You Love That Pays You Well

<written by a human being>

Desktop software licensing, it turns out, is also a whole story. There are different options for how licenses can work. I'm selecting options for my application, which should work completely offline. And one of the options that's clear to me is online activation with a request to an external license verification service. To verify the license, you can build a worker that will query a database of generated keys and check for their presence. Such a worker can be set up, for example, on the basis of Cloudflare Workers. This is, of course, simplified for understanding, and a lot needs to be added for everything to work reliably and securely. However, there are also offline activation and licensing schemes that are based on cryptography, hardware protection keys, and time-based licenses whose validity is checked locally without connecting to the internet.

<written by a human being>

When you work with a web application, an obligatory part is its cybersecurity. Because, of course, you want to protect the server on which the application runs, all the data that's in it from attacks by malicious actors. When you develop offline desktop software, it's necessary to protect it from decompilation or, in simple terms, disassembly where you can literally view the source code and also access the database itself. Of course, 100 percent protection from this doesn't exist, but the following mechanisms are applied. For example, obfuscation of program code, which makes its analysis difficult. Further, compilers that make decompilation or its reverse engineering difficult. Database encryption. There's also a technology stack that's initially more resistant to decompilation. When choosing architecture, all of this is also extremely important to consider.

<written by a human being>

For some reason we've forgotten about good old offline desktop applications that work without the internet, that work fast, that don't require a network connection to load something or to generally work normally and fully. We have internet almost always up to 100 percent of the time, but the charm of the speed and clarity of response to your actions of full-fledged offline desktop applications that don't require downloading anything from the network - this is actually a different level.

The fastest way to become irreplaceable is to understand that your uniqueness already exists Read more about how to Red Pill Your Career: From Replaceable Employee To Irreplaceable Creator

<written by a human being>

A curious fact about many modern "desktop" applications: I'm writing this intentionally in quotes because I want to say that they are essentially not desktop applications in the primary understanding, but use embedded browser engines that render a web interface. This is done so that it's easier for developers to write one front-end that's reused everywhere, regardless of the system where it's running: whether you're using the software application in a browser or opening a native application. For example, ChatGPT or Claude applications are built on the same principle and display a loaded web interface using the same principle as when loading internet pages. Such applications, naturally, won't work fully offline. At the same time, by the way, they use a local database that stores the client state, but there's no point in storing the dynamic data necessary for the main business logic, since it's loaded in online mode. Overall, there are a huge number of examples of such applications that enjoy great popularity. For example, Steam and Discord are also built on the basis of web interfaces, rather than as desktop applications in the original sense. Although Steam's backend is still hardcore C++.