Posts of Mike

rate the UI

I went back through my prompts from the last few projects and noticed something uncomfortable. Most of them weren’t about building new things. They were about re-explaining things I had already built: the architecture, the conventions, which feature depended on what, and where the last session stopped and scattered build plans. So I built a companion skill for Claude Code and Codex that generates this context from your repo and keeps it current as you work and I have used it for a while so I'm open-sourcing it. Your agent reads it before every task and writes to it after. You explain your project once. From then on, your prompts are about what’s next, not what already exists. Coming soon → https://contextual.mikaelen.site

No text

I hate AI slop websites like makes me hate the vibecoder behind also 😭

I have one idea 💡 Imma ship it today also a challenge for me Cool shit trust me

Its time to get a vps fr minimum 2k per month and 12 gb ram cloud vps this is crazy good offer

No text

I feel weak when i cant max out my limits 😭 why are they generous now I'm running fable with xhigh effort and a bil tokens per day minimum and it stays there every time I visit it is still 10% or something smaller

Training a model for a community project 🤯 it is the new standard y'all

Beautiful morning 🌅 The sky is clear once again

A cracked guy made this 🔥

No text

No text

No text

TG Brain is live I mapped Ethiopian Telegram creators as a living 3D brain. Neuron by Neuron, then trained a model with all the data that exists. Every connection is real. It comes from who mentions whom, who posts about the same things, and how ideas have actually moved between channels. What you can do: Find yourself and see your neuron light up Trace the pathway between any two creators Test an idea and watch it spread. See who’d pick it up first, who’d pass it on, and why each one was chosen Share your neuron to your channel There is No AI guessing. The model learns only from what’s actually been posted. Go find your neuron 👇 https://tg-brain.pro.et/ If you’re in it, share yours. That’s how the brain grows

Rust is the future

In the age of AI, technology by itself is becoming less defensible. Code, interfaces, workflows, and even fairly sophisticated product features can now be reproduced much faster than before. If the main reason your company is difficult to compete with is simply that your software took a long time to build, that advantage is becoming weaker every year. The important shift is that implementation is being commoditized. When everyone has access to powerful models, coding agents, cheap infrastructure, and open-source components, the advantage moves away from who can build the feature and toward who can learn what should be built next. That is what I mean when I say speed is the moat. I do not mean raw coding speed, and I do not mean shipping random features faster than everyone else. I mean the speed of the entire learning loop: how quickly you can notice a change in the market, form a hypothesis, build something, distribute it, observe how real users respond, kill what does not work, update your understanding, and make the next decision. The company that completes that loop faster gets more attempts at discovering the truth. Imagine two companies attacking the same market with access to roughly the same technology. One company can run ten meaningful experiments in the time the other runs two. Even if most of those experiments fail, the first company has interacted with reality five times more often. It has learned more about customer behavior, willingness to pay, distribution, onboarding, retention, messaging, product priorities, and where its assumptions were wrong. Over one month the difference may look small. Over several years it compounds into a completely different understanding of the market. This is why copying a product does not necessarily mean catching the company that built it. A competitor can often see your interface, reverse engineer your features, copy your pricing, and reproduce large parts of your product. What they cannot see as easily is everything that produced those decisions. They do not automatically have the customer conversations that caused you to remove a feature, the failed experiments that taught you which segment not to target, the data that changed your onboarding flow, or the insight that caused you to start building the next product before the current one became obvious to everyone else. They can copy the visible output, but they cannot instantly copy the learning history behind it. The real question therefore is not whether someone can copy your product. In modern software, the answer should often be assumed to be yes. The more important question is whether copying today's product is enough to catch the company. If you are learning and evolving significantly faster than your competitors, they may reproduce where you were without understanding where you are going. They are copying a snapshot while you are operating a moving system. AI makes this dynamic even more important because it lowers the cost of experimentation. If building an experiment used to take a month and now takes two days, the value of being able to choose, launch, and evaluate the right experiments becomes much greater. The bottleneck moves from engineering capacity to organizational intelligence. Judgment, product taste, customer proximity, prioritization, and decision speed become more important precisely because the underlying implementation becomes easier. This also explains why small companies can sometimes outperform much larger incumbents. Large companies may have more engineers, more capital, more distribution, stronger brands, and deeper relationships. Those are real advantages. But scale often introduces coordination costs. A change that a founder-led team can make in a day may require meetings, planning cycles, cross-functional alignment, legal review, budget approval, and organizational consensus inside a large company. None of those processes are irrational; many exist for good reasons. But they create latency between seeing something and acting on it. A startup's advantage is therefore not simply that it is smaller. It is that it can compress the distance between information and action. A user complains in the morning, the team understands the underlying problem by noon, a new version ships that evening, and the next day the team is already measuring whether the change worked. If that cycle happens continuously, the startup can accumulate hundreds of small learning advantages before a slower competitor completes a few major planning cycles. However, speed should not be confused with permanent defensibility. Speed itself can eventually be copied. Competitors can hire better teams, reorganize, use the same AI tools, or create faster internal units. The deeper value of speed is what it allows you to build before that happens. Speed buys time, and inside that time you can accumulate the harder forms of defensibility: proprietary data, customer trust, brand, workflow integration, distribution, switching costs, ecosystem relationships, network effects, and operational knowledge. This distinction matters because those advantages usually do not appear automatically at the beginning of a company. A startup rarely launches on day one with massive switching costs or a dominant network effect. Those things are earned through repeated interaction with the market. Speed increases the number of chances you have to earn them before someone else does. It is also important to separate distribution from network effects. Having access to customers, investors, buyers, sellers, or influential people is valuable, but access alone is not necessarily a moat. A true network effect exists when each additional participant makes the product more valuable to the other participants, making the network increasingly difficult to replace. A competitor with better relationships may be able to reach your customers, but that does not automatically mean they can recreate a product whose value depends on the existing network itself. At the same time, there are situations where speed is not enough. If an incumbent already has extremely strong network effects, regulatory protection, massive economies of scale, exclusive distribution, or very high switching costs, simply moving faster may not overcome those advantages. That does not weaken the argument. It clarifies it. Speed is most powerful as the mechanism that allows an emerging company to discover an opening, exploit it quickly, and build durable advantages before the market closes around it. So the strongest version of “speed is the moat” is not that every other moat is fake. It is that in an environment where the cost of building software is collapsing, speed of learning becomes the foundational advantage from which many of the other moats are created. Your code can be copied. Your interface can be copied. Your feature list can be copied. Your pricing can be copied. Your AI stack can be copied. What is much harder to copy is a company that has spent years learning faster than everyone else and has converted that learning into customers, data, trust, distribution, workflow ownership, and network effects. That is the real argument. Speed does not make you impossible to copy. Speed makes copying you insufficient.

I wanna explain this to you even more clearly because you should not be led by the wave

It might take you a few to understand this till you are on the real battleground

In the age of AI, your tech is not a moat. If someone can rebuild your product in a few prompts, it was never a moat, it was a weekend project. The only thing that can’t be cloned in a weekend is how fast you move: how quickly you ship, validate, kill what doesn’t work, and jump to the next bet while everyone else is still in a planning meeting. Claude Code made building cheap for everyone, so the winners are the ones who run 10 experiments while others run 2, and learn something real from every one. Incumbents need a roadmap and three approvals to ship a button. You ship it by lunch. Speed is the moat, and everything else (users, trust, data) gets built inside the window it buys you.

Speed is the only moat

Older sister's are a blessing

Trivium Online is fully out now see you Thursday! And for our users who joined the wait list join now and get ready for exams. https://www.triviumonline.com.et

the reason I follow doughnut 👌🏻 this is going to change alot

One tool i am absolutely in love with recently is openbot. Everyone loves grokbot or dots but this lets use multiple subs and is just realllyyy nice when coordinating things as well. https://openbot.run/ #Resources

No text

No text

No text

No text

No text