What's Next for Software Engineers?

TL;DR Writing code is becoming another layer of abstraction, just like machine code gave way to C, and C gave way to Python. English is the new programming language. The moat isn't code anymore. It's ideas, reputation, trust, and the wisdom to build a life that doesn't depend on one employer.

Hey Software Engineer! I know why you're here. You've seen what Opus 4.6 and GPT 5.4 can do. You've watched AI write production code, not autocomplete suggestions, not boilerplate snippets, real working software. And you're wondering: what happens to me?

You're not alone. Every generation of programming went through this. Machine code. Assembly. COBOL. Pascal. C. Java. Python. Each was a layer of abstraction over the one before it. Each time, the old guard said it wasn't real programming. Each time, more people got access to building software. More value got created. We're watching the next jump happen right now.

AI-assisted Development Environments (ADEs) are the new IDEs. English is the new programming language. The cost of writing code is approaching zero.

Here's what I've learned about staying relevant.

1. AI as Your Daily Multiplier

Before anything else, use the tools. If you're not using AI to multiply your output right now, you're leaving 5-10x productivity on the table.

• Coding agents. Claude Code, OpenCode (which works with your existing GitHub Copilot subscription), Codex. There are more advanced ADEs like Goose and others emerging every month. Pick one. Use it daily. Get comfortable giving it real tasks, not toy problems. Build workflows that works for you.

• Grounding. LLMs hallucinate. They'll confidently give you deprecated APIs and functions that don't exist. Use Context7 as an MCP to feed your LLM version-specific, real-time documentation. Use web search tools to ground your LLM with up-to-date information beyond code docs.

• Persona. Use skills and system prompts to make your LLM work the way you think. Want it to plan with you before writing code? Use something like superpowers /brainstorm. Want tracked requirements that evolve over time? Use SpecKit or OpenSpec. The raw model is a starting point. The configured model is the tool.

• Multi-agent. Don't stop at one agent. Combine system prompts, skills, and specialized agents that handle different parts of your workflow. One agent for planning. One for implementation. One for review. This isn't science fiction. People are doing this right now.

• Testing. Use Playwright to have AI navigate your website, interact with it, and let the LLM find UX issues. Let machines test what machines built.

• Design. Generate HTML mocks. Iterate on them. Convert to Figma. Then feed the Figma designs back to your coding agents for pixel-accurate implementation. The loop from idea to working UI has never been shorter.

• Open source models. Run models locally on your Mac using LM Studio or vLLM. Kimi K2.5, GLM-5, Qwen 3.5, and MiniMax 2.5 are all strong right now. Playing with these teaches you how LLMs actually work, that you send all conversation turns every time, how context windows behave, what temperature and top-p do in practice. That understanding opens you up to different ways of configuring and deploying models that most engineers never think about.

• T-shaped. When AI writes the code, being a specialist in one language or framework isn't enough. Understand products. Suggest UX improvements. Build scalable infrastructure. Architect systems end to end. The engineers who thrive will be the ones who can move across the stack, think about the user, and make decisions that go beyond implementation. You can't just focus on one thing anymore.

• Stay current. Follow engineers who are riding this wave. Karpathy (nanochat, autoresearch), Boris Cherny (thread), Peter Steinberger (blog, Lex Fridman interview), Gary Tan (check out gstack). Not because they're authorities you should blindly follow, but because watching how they structure their setups teaches you what's possible. Synthesize it. Figure out what works for you. Stay plugged into X, Reddit, and LinkedIn. That's where real pain points, real solutions, and real innovation get shared daily.

2. Build Your Reputation

Making yourself visible is no longer optional.

• Home base. Build your own website. Route traffic through a blog. Write about whatever you think is worth sharing. It doesn't need to be groundbreaking. It needs to be yours.

• Open source. Nothing is too small. The entire ecosystem we stand on exists because someone decided their small thing was worth sharing. Linux started as one student's kernel. C was designed by two people at Bell Labs. Git was built because Linus needed a better version control tool. Every one of these was "too small" at some point. Your project might be a CLI tool that saves you 10 minutes a day. A library that solves one problem well. A template that helps others get started faster. Ship it. Someone out there needs exactly what you built.

• Write. Put your thoughts where the right audience will find them. That might be your blog, a subreddit, a Hacker News post, or a thread on X. The medium matters less than the consistency. Show up. Share what you know. Build a body of work that speaks for you when you're not in the room.

3. Build Something for Yourself

• Execute. Had an idea for years? This is the time. The resources at your disposal are unprecedented. Use LLMs to research your market. Talk through your idea with an AI partner. Let it poke holes in your assumptions. Build specs. Then let it build the application. Write tests, good ones, that actually catch regressions.

• Market. The public is welcoming good ideas. Look at what happened with ClaudeBot/MoltBot/OpenClaw and MoltBook. Not just that, there are more examples everywhere: WisprFlow, ElevenLabs, HuggingFace, Manus, Scale.AI, Suno, Krea, NotebookLLM, Pika. Small teams and solo builders are shipping products that compete with companies 100x their size.

• Leverage. SaaS isn't dead. As long as you innovate, stay ahead, price accurately, and build trust, you will thrive. More than ever. Because we as humans want leverage. We always have. We pay others to handle complexity so we can focus on what matters to us. That fundamental desire doesn't change because AI exists.

• Reliability. The barrier to building has collapsed. The barrier to building something good hasn't. Uptime. Quality. Security. Support. Leverage platforms that handle the heavy lifting: Supabase for your backend, Clerk or Auth0 for identity, Railway or Render for deployment, Vercel for frontend hosting, Resend for email, Stripe for payments. These tools let a solo builder deliver the same reliability that used to require a team of twenty. Use them. Ship fast, but ship something you can stand behind.

• Distribution. Learn the basics of SEO and GEO (Generative Engine Optimization). Understand how search engines and AI models surface content. Use web analytics to see what's working and double down on it. You don't need to become a marketing expert, but knowing how to get your work in front of the right people is a skill that compounds. Writing great code used to be enough. Now you have to make sure people can find it.

4. Ideas Are the New Moat

• Speed. Generating ideas, hardening them with an AI partner, and quickly putting together a working demo is going to be what separates you from the rest. Ideas that reduce costs, improve efficiency, and increase build quality are going to shine. Not because they're novel in concept, but because for the first time you can go from idea to working prototype in days instead of months.

• Explore. Blockchain. Peer-to-peer networks. Decentralized systems. Netflix renders video through a P2P network mesh. Most people don't know that. You'll only land on an amazing idea when you know what's possible. You can only connect the dots when you've collected enough dots. So collect them. Read widely. Experiment. Try things that feel unfamiliar.

• Persistence. Don't stop if someone says it won't work. Build it and see for yourself. At worst, you walk away with good learnings. At best, you prove everyone wrong.

• Reach out. Cold email whoever you want to. Within your organization or outside. You never know what others are looking for. Have an idea and want to take it to the next level? Send genuine cold emails. Be specific about what you've built and why it matters. You have a shot. More of a shot than you think.

5. Have a Backup

No matter what you do, protect yourself.

• Finances. Your career depends on a company whose financials you can't control. If there's hardship, they will let you go. Keep 6-12 months of runway. Always.

• Skills. Stay interview-ready. Keep your system design sharp. Keep your fundamentals current. The laid-back days are gone.

• Diversify. If possible, build a business that is local to your region. Something simple that gives people back their time so they're willing to pay for it. Or something that doesn't just involve software. Not everything needs to scale. Some things just need to work.

• Connections. As we enter whatever comes next, abundance or the opposite, the only way you can survive is because of your family, friends, and connections. They can get you a job. Share your hardships. Suggest something that would work great for you. Cherish your humans.

• Generosity. Be helpful to others. You never know what comes back.

The Timeline

The pace of innovation is getting harder to keep up with every day. But enterprises move slow. They take time to adopt, time to pivot, time to retrain. That lag is the buffer. I believe we have 5-7 years for this wave to reach every corner of the industry, and when it does, the workforce shrinks to maybe 30-40% of what we have today. Not because engineers aren't needed, but because each engineer's output multiplies so dramatically that fewer people produce the same results. The cost of writing code goes to zero. People will build a lot of new products. Trust is what's going to be the differentiator.

I believe there are and will be strong opportunities in cybersecurity, privacy (apps that promote privacy and E2E encryption, inherently trustworthy by design), blockchain, peer-to-peer networks, self-hosting, IoT, the home assistant space, generative gaming, and augmented reality. Not just in any single domain, but at the intersection of these domains. Privacy meets IoT. Blockchain meets P2P. Generative gaming meets AR. The most interesting problems live where fields overlap, and that's where the most valuable solutions will come from. There will also be strong demand for consultants who can help migrate traditional web apps to agentic-first applications. These jobs are hot right now and will stay that way as every company figures out how to rebuild their products around AI. Another stream of opportunity is applying this technology to domains outside software: physics, chemistry, molecular biology, biotech, waste management. The engineers who can bridge AI and a domain they care about will be in a category of their own.

I believe three categories of jobs will define the future:

1. Scientists. The people building cutting-edge LLM models, inventing new architectures, advancing distributed training, building efficient inference infrastructure, and pushing GPU/TPU hardware acceleration forward.
2. Builders. Agentic AI consultants who take that technology, hardware and software, and apply it to advance different fields, transform existing products, and help enterprises stand up local LLM inference infrastructure.
3. Creators. AI influencers and generative content creators who can attract eyeballs, shape narratives, and build audiences around what this technology makes possible.

Figure out which category fits you and go deep.

If you're in this industry by passion, you'll figure out your niche and thrive. The tools have never been better. The opportunities have never been wider. The only question is whether you adapt or wait.

Best of luck, mate.

Thoughts? Hit me up at [email protected]