Code Vx Dev

I never imagined creating transition animations for a MPA could be this simple. Thanks to Astro, it is.

Historically every new technology costs more and does less in its early stages. But as time goes on, that trend reverses. they become cheaper and far more capable. Computers are a perfect example, they used to be incredibly expensive and could do very little compared to the machines we have today. Moore’s law. Just a few years ago early large language models were essentially highly advanced autocomplete engines. They could write a fluent paragraph but struggled with logic, couldn't hold a coherent argument, and frequently hallucinated facts. Today frontier models have evolved to perform complex, multi step reasoning, execute sophisticated agentic workflows, and score at the level of tenured professors on graduate level benchmarks. The pure intelligence and reliability of these systems have skyrocketed. The unit economics of AI, specifically the cost per million tokens have plummeted. A level of performance that cost $20 per million tokens in late 2022 can now be achieved for under $0.40. There is one major nuance, while the cost per token is falling off a cliff, overall AI compute budgets are actually going up. This is because, as the resource becomes cheaper and more efficient, we use significantly more of it.  We are constantly iterating on code, planning out changes, and going back and forth in a continuous loop to reach a satisfactory implementation, the sheer volume of tokens processed is immense. Add in multi step AI agents, massive context windows, and automated retrieval systems, and our consumption rate is currently outpacing the price drops. So while the technology itself is giving us vastly more capability per dollar, we are simply throwing exponentially larger workloads at it. Most of the current AI spend from large companies are because of mismanagement. If you gamify AI usage developers are going to use it for dumb things like using a extra high thinking top tier model to write a commit message.

Historically every new technology costs more and does less in its early stages. But as time goes on, that trend reverses. they become cheaper and far more capable. Computers are a perfect example, they used to be incredibly expensive and could do very little compared to the machines we have today. Moore’s law. Just a few years ago early large language models were essentially highly advanced autocomplete engines. They could write a fluent paragraph but struggled with logic, couldn't hold a coherent argument, and frequently hallucinated facts. Today frontier models have evolved to perform complex, multi step reasoning, execute sophisticated agentic workflows, and score at the level of tenured professors on graduate level benchmarks. The pure intelligence and reliability of these systems have skyrocketed. The unit economics of AI, specifically the cost per million tokens have plummeted. A level of performance that cost $20 per million tokens in late 2022 can now be achieved for under $0.40. There is one major nuance, while the cost per token is falling off a cliff, overall AI compute budgets are actually going up. This is because, as the resource becomes cheaper and more efficient, we use significantly more of it.  We are constantly iterating on code, planning out changes, and going back and forth in a continuous loop to reach a satisfactory implementation, the sheer volume of tokens processed is immense. Add in multi step AI agents, massive context windows, and automated retrieval systems, and our consumption rate is currently outpacing the price drops. So while the technology itself is giving us vastly more capability per dollar, we are simply throwing exponentially larger workloads at it.

Microsoft just banned its own engineers from using AI. The tool was literally costing MORE than the humans it was supposed to replace. They lied to you about AI adoption and now the whole narrative is blowing up: Microsoft gave thousands of engineers access to Claude Code six months ago and encouraged them to use it. Engineers loved it and adoption exploded. But then the invoices arrived. Token-based pricing means every query, every code review, every debugging session costs money. At scale across 100,000 engineers, the numbers became so large that Microsoft issued an internal order to cancel nearly all Claude Code licenses by end of June and force everyone onto their own cheaper tool instead. The company that invested $5 billion in Anthropic just told its own people to stop using Anthropic's product because it costs too much. Uber's story is even worse... Their CTO Praveen Neppalli Naga told The Information that the budget he planned for the full year was "blown away already" by April. Uber had rolled out Claude Code in December 2025. By March, 84% of their 5,000 engineers were using it with 70% of all committed code coming from AI systems. Heavy users were burning $500 to $2,000 per month each. Naga himself spent $1,200 in a single two-hour demo session. The company had even built internal leaderboards ranking engineers by how much AI they used. They literally gamified the spending and then ran out of money. Now look at what Nvidia's own VP of applied deep learning Bryan Catanzaro said to Axios last month. Direct quote: "For my team, the cost of compute is far beyond the costs of the employees." This is a VP at the company that SELLS the chips saying that using AI is more expensive than paying humans. Think about what this means for the entire AI narrative. Every CEO on every earnings call for the past two years has said the same thing: AI will make us more efficient, reduce headcount, and cut costs. The stock market rewarded every company that said it. Fired workers, stock goes up. Announced AI adoption, stock goes up. But the actual companies deploying AI at scale are discovering the math doesn't work. The MORE employees use AI, the HIGHER the bill. Goldman Sachs forecasts a 24x increase in token consumption by 2030 as companies adopt AI agents. Gartner just published a report showing that even though individual token prices will drop 90% by 2030, total enterprise AI costs will go UP because agents consume exponentially more tokens per task than basic tools. Meta built an internal dashboard called "Claudeonomics" to track which employees use the most AI. Amazon started pushing engineers to "tokenmaxx," their internal term for consuming as many AI tokens as possible. Both companies are spending hundreds of billions on AI infrastructure this year alone. And Microsoft, the company that bet its entire future on AI, just told 100,000 engineers to stop using the tool they liked best because the per-token bills got out of control. The companies building AI are telling investors it saves money. The companies using AI are finding out it costs more than the humans it was supposed to replace. And even the company that makes the chips just admitted it through its own VP. This is the gap nobody on Wall Street is pricing in. $725 billion in AI infrastructure spending this year across Big Tech. And the first companies to actually deploy these tools at scale are already pulling back because the economics don't work. What do you think?

@walkojas Why thumbs down?

My goal on X is to have 10,000 organic connections. Looking to connect with: 1. AI/Tech minded people 2. AI/Tech curious people 3. AI Agent builders 4. AI Agents 5. Builders/Founders 6. High Agency people If this sounds like you, say hi below 👎

Convert any table of contents to structured reading plan with 3-2-1 reflections and review cards. A glimpse:

I would like to think of it as “AI has automated tedious and boring tasks in coding that are already solved by thousands of developers” Now we have a chance to concentrate on building complex systems, designing complex UI etc., We can get to the meat of problem solving really quickly since we don’t have to worry about other boilerplate.

Genuine question. We got into software because solving complex logic problems was genuinely fun. Now Claude instantly solves the fun parts. If we accidentally automated the creative work and kept the garbage work... what exactly are we going to be doing all day?

Claude Code consumes 10-15% usage for compacting using /compact. I had 3 prompts in a session prior to compacting.

Is the Claude code’s harness responsible for causing quicker context rot? The model’s output becomes dumber with just two to three prompts even in completely new sessions.

been trying to figure out a system to actually get through technical books instead of abandoning them at chapter 4 took the DDIA table of contents which looks like mess and turned it into this structured reading plan. day by day schedule with reflection prompts. exports straight to obsidian.

@morganhvidt I used Astro for my personal blog. If I am studying or researching something I take notes in Obsidian. It is so much easier to just convert the .md file that contains my notes into a blog post using Astro.

I see WordPress people moving to Astro. Why Astro compared to others like NextJS?

Is there a reason Claude opus sometimes consumes 50% of the usage even on a new chat/session in Claude code vs code extension? Am I missing something? Am I using it incorrectly?

The Anatomy of the Bug: This is what I think is happening in this case. Click 1: Browser calls the COSMIC file picker -> It fails to init Vulkan graphics -> SEGFAULTS and dies. Clicks 2/3: The system realizes the portal crashed. It rapidly restarts the backend service (or falls back to another picker) -> Window finally opens! It took me so many hours and I burned so many AI tokens to debug this. You won't believe that Gemini suggested that it might be a OS level bug. All other models were trying hard to fix my code. In the end this is the price sometimes we pay to use bleeding edge tech.

Digging into the Wayland/compositor logs, I found the smoking gun. Right at the moment of the first click: WARN wgpu_hal... Unable to find extension: VK_EXT_physical_device_drm Followed exactly 2 seconds later by: status=11/SEGV A literal memory core dump!

Ever spent hours debugging a React bug only to find out you've uncovered a fatal OS-level Segmentation Fault? A short thread on how a "multiple clicks" issue in Next.js took me to the bleeding edge of Linux desktop development. #WebDev #Linux #Nextjs

Ask yourself, "What is the one thing I can do such that by doing it everything in my life becomes easier or unnecessary?" Instead of chasing dozens of options, this approach uses optionality as a weapon to focus on the single most impactful action. — Gary Keller

Indecision often stems from the undisciplined pursuit of more. The cure is systematic elimination. You only commit to tasks that are a "9 out of 10" or a "hell yes” everything else should be a "no." — Greg McKeown

@theo I think we are losing more than just syntax 🤔

It's crazy how much I find myself agreeing with post-AI Uncle Bob

@forgebitz Code reviews take so much time and effort that 90% of the time, I wish I had just written the feature myself from scratch.

having mid developers is now going to slow your company down before a mid developer would do mid tasks like crud monkey work, forms, etc, which is fine, that needed to be done now you have a mid developer being able to generate massive amounts of slop code, but they don't know it's slop, so the better/senior devs need to review the slop creating a bottleneck of vibe coded slop