Security Response Engineer, Incident Response

Not Another AI Panic Post but Just What’s Actually Happening in AI Right Now

This is a long post, read it only if you genuinely care about your job. If the looming AI storm stresses you out, feel free to skip it. I’m a seasoned AI engineering manager based in San Francisco. I’ve been deep in this space for over a few years. Built systems, led teams, scaled products. But what I’m seeing out here right now? t’s different. It’s bigger. And most people have no clue how fast it’s coming for their jobs. Everywhere I go - cafes, meetups, casual hangs, people won’t shut up about AI agents. Everyone’s out here tryna build the next big GenAI startup like it’s a freakin’ gold rush. Ain’t nobody tryna code anymore, they’re just wiring up agents to do all the dev work for them. Full-blown software pipelines, DevOps, SWE, QA, Data, Security, getting bodied. Web dev? Already smoked. That shit’s over. Just chaining LLMs to ship products faster than full engineering teams ever could. This is the exact reason why you see many posts about the US job market collapse and not encouraging students to go to the US for masters, all those fancy cost cutting covers are just bullshit. Software engineering as we know it is being automated from the top-down. Entry-level jobs? Already gone, Mid-level roles? Getting phased out by fine-tuned agents, Even senior engineers? Only safe if they’re building the AI that’s replacing everyone else. Just as you know Sam Altman tweeted saying they're short of GPUs, and last month they collaborated with the US govt, Cisco, NVIDIA and building 500 billion dollars worth of AI infrastructure. OpenAI’s Chief Product Officer straight up said they’re focused on building reasoning models, Models that can think through a prompt for months. Let that cook for a second. The kind of reasoning models that don’t just answer questions, they solve problems you can’t even define yet. So if you’re sittin’ around chillin’ on your React app or tuning Jenkins pipelines like it’s 2019, wake the fuck up. I don't even know what to suggest for SWs but at least Learn how these models work. Dive into LLMs. Understand token limits, context windows, prompt engineering, reasoning trees. Hell, build an agent. Or at least know what one is before you get tossed. Agents don’t sleep. They don’t quit. They don’t ask for raises. Your job description just became a prompt template. And You’re not getting replaced by someone cheaper - you’re getting replaced by something smarter. Let that sink in And that’s the nicest way I can put it. I'm not saying this is happening tomorrow - this is in the next 3-4 years, I'm even scared if I'll be able to save my job it's that scary if you truly understand & worked on AI. Anyone disagreeing with AI not taking jobs - you probably don't know what AI's brother. We are so sorry and good luck. And that’s the nicest way I can put it, The truth is stop glorifying your jobs. Stack some cash. Hard times are coming for all of us. Especially if you're in that 25–35 range, just starting out or juggling family expenses, consider this a wake-up call. Spend wisely. No one should have to face supporting a family while jobless or even on the edge of losing one. GOD BLESS ALL OF US.

Has anyone tried throwing raw Indian addresses at Claude 4.7 Sonnet yet?

This is a quick post read it only if you genuinely care about where logistics tech is heading. If the looming AI storm stresses you out, feel free to skip it. I’m a seasoned engineering lead handling ops data for a last-mile delivery startup. I’ve been deep in this space for over a few years. Built routing engines, scaled tier-2 city ops. But what I just saw this Tuesday morning? t's different. It's bigger. Has anyone tried throwing raw Indian addresses at Claude 4.7 Sonnet yet? I just used Cursor to write a quick parser for our tier-2 city ops data, and it straight up smoked everything we've built in the last three years. Everywhere I go - planning meetings, tech huddles, people won't shut up about writing better regex or hiring more data validation folks. Ain't nobody tryna understand that the game has changed. Indian addresses are notorious. "Near the broken water tank, opposite Sharma sweets, 3rd gali" - traditional geocoding getting bodied by this. We used to have full-blown teams just cleaning and standardizing this garbage. That shit's over. Just chaining the new Sonnet model to parse and output clean JSON coordinates faster than full engineering teams ever could. Entry-level data ops? Already gone, Mid-level mapping engineers? Getting phased out by fine-tuned agents, Even senior backend guys? Only safe if they’re the ones wiring up the API. So if you’re sittin’ around chillin’ tuning your Google Maps API fallbacks like it’s 2024, wake the fuck up. Dive into LLMs. Understand context windows and how these reasoning models can actually comprehend localized Indian spatial chaos. The model doesn't just extract text, it solves the routing problem you can't even define yet. Your job description just became a prompt template. And You’re not getting replaced by a cheaper third-party vendor - you’re getting replaced by something smarter. Let that sink in And that’s the nicest way I can put it. The truth is stop glorifying your complex microservices for simple ops problems. Stack some cash. Learn to build agents. GOD BLESS ALL OF US.