Pratham
@Prathkum
I talk about web, AI, API, and social • Building experiences at @APILayer • Prev @Rapid_API @HyperspaceAI
AI will create a lot of jobs. when internet was new, it disrupted a lot of jobs like media, travel agents, printing press etc. but it also created a lot of new roles like web developers, seo, cloud engineers, growth teams, devops, social media, etc. none of these jobs really existed before the web became the default layer of the world. similar thing happened when mobile devices became popular. it created a ton of new jobs like ios developers, network engineers, ux, etc. AI is the new technological shift. writing a software is easy now. anyone can generate code, design frontend, and ship a prototype in hours. but that is exactly the point: when building becomes cheap, something else becomes hard. most AI apps never reach meaningful scale. distribution, integration, workflows, reliability, and real user value are hard problems. i see two relatively new roles are already getting popularity. 1/ gtm engineer to make sure what is built in a day stays for a long. making sure it reaches users and grows. 2/ forward deployed engineer: as most companies want to enable AI/agent in their products but not a lot of engineers are fully educated about the core tech of it. so next time don’t worry about the AI will take your job. worry about whether you are competitive and educated enough to survive the new wave of roles.
The way we write code has changed but not the way we engineer software. We use AI to generate the code, we use better tools, and we move much faster. Great, we got code generation on demand. But engineering is the same as it was a decade ago. • managing complexity • understanding the system • scalability • security and vulnerability Today, anyone can easily build a todo app for themselves. In their favorite colors, functionality, the way they want it. That's coding. But taking that todo app to 1000 users without breaking it. Without leaking users' data. Managing DBs. That's engineering.
We spent twenty years optimizing how to write software. The next twenty might be about deciding what software deserves to be written at all. For the last decade, we cared about clean abstractions, reusable components, proper architecture, modularity, maintainability. All of that made sense because humans were the primary producers and maintainers of code. The constraints of our brains shaped the structure of our systems. But something subtle is changing. When I open my editor today, I’m not staring at a blank file anymore. I’m describing intent. I’m reacting to output. I’m refining drafts. The first version of almost anything is now generated in seconds. The role of effort has shifted from construction to evaluation. This changes the unit of leverage. Earlier, leverage came from knowing how to implement something efficiently. Now leverage increasingly comes from knowing what should exist in the first place. If you can articulate the problem well, the implementation follows almost automatically. And this has second-order effects on SaaS. A lot of products historically justified their existence by abstracting away complexity that was painful for humans to build repeatedly. Internal dashboards, workflow tools, small CRMs, reporting layers. These made sense because custom-building them required weeks of engineering time. But if describing your workflow in plain English can generate a bespoke internal tool in minutes, the economics start to wobble. The advantage of “we already built it for you” weakens when building it yourself is no longer expensive. I don’t think SaaS will disappear. I think the definition of value moves up the stack. Distribution becomes more important than implementation. Data becomes more important than UI. Trust becomes more important than feature lists.
MiniMax M2.5 gives you Opus 4.6 speed and performance. It’s roughly 10 times cheaper. And Open Source. I am on the coding plus plan, which gives me 300 prompts per 5 hours. Check out the API here: http://platform.minimax.io/docs/coding-p… Some use cases: 1. Build games
MiniMax M2.5 > Claude Opus 4.6 I have been using it for a couple of days and have never hit the limit. I am currently on the Coding Plus plan which gives me 300 prompts per 5 hours. My new setup is OpenCode + MiniMax M2.5 What I have noticed: M2.5 – handles multi-step coding workflows and repo-level reasoning more reliably – better at chaining actions. think: generate test, fix bug, refactor, repeat – high TPS throughput M2.5 hits ~80%+ on SWE-Bench coding benchmarks on par with Opus and runs these tasks 37% faster than its predecessor while costing ~1/10th - 1/20th the price. Check out the video.
My take on Block layoffs (my own): The reason isn’t AI getting too smart and that’s why they don’t need humans anymore. The reason is their bloated workforce. 10,000 employees in a company was never justifiable to me, at least to me (I might get a lot of hate for saying that). If you hire 10,000 people for your company, you are not making their lives easier. You are just putting them in a spot where they would be under constant threat of getting laid off. Jack is an awesome mind. I respect him for his crazy ideas. But he has a proven record of running a bloated company. I’m talking about Twitter. Before Elon acquired it, Twitter didn’t have any major feature additions (at least I don’t remember), Twitter was a bloated workforce company. But after Elon acquired it, Twitter became X, there was a massive workforce cut, a lot of features were added, and it grew overall. Saying AI impacted 4,000 jobs at Block isn’t 100% accurate. The only way to save yourself: enable AI in your life, try not to join bloated companies, build something with AI. So stop blaming AI.