Every first day of the month, Hacker News publishes its “Who Is Hiring?” thread. It’s one of the most honest signals of the job market I know — without the filter of recruiters, without the theater of LinkedIn, just founders and engineers posting what they actually need. I’ve been reading these threads for years. The April 2026 one just dropped a few hours ago, and combined with the 482 comments from March, the picture is clear enough to act on.
These are the five patterns I’m seeing.
1. “AI-native” is no longer a differentiator — it’s the starting point
The companies posting in this thread aren’t adding an AI feature to an existing product. They are the AI. Shepherd just announced a $42 million Series B to build what they call “fully autonomous underwriting” for commercial insurance. Softmax — founded by OpenAI’s former interim CEO — is hiring engineers to build multi-agent AI environments focused on alignment research. BIT Capital in Berlin is betting its entire investment infrastructure on AI to hit a target annual return above 30%.
These aren’t “we integrated ChatGPT” companies. They’re building the infrastructure layer of the AI economy. The distinction matters: working at an AI-native company means you’re building the primitive, not the product on top of it.
2. The agentic stack has consolidated
Eighteen months ago, “agentic AI” was a conference buzzword. Today it appears in job descriptions with surgical precision. BIT Capital explicitly lists agentic scaffolding, RAG, and LLMOps as part of their infrastructure role scope. PrairieLearn is building LLM agents for content creation and vision-language models for automated grading. An AI platform for credit unions is looking for someone to own “agent orchestration and tool calling” from day one.
The stack converged. If you’re serious about this market, the combination of RAG, vector databases (pgvector appears repeatedly), agent orchestration, and LLMOps tooling is the new “knowing Linux and TCP/IP”. It’s not optional context — it’s the work itself.
3. TypeScript won
I don’t say this lightly, but the evidence in the March and April threads is compelling. Practically every startup in the thread runs some variant of: TypeScript + Node.js + React + PostgreSQL. Schema Labs, PrairieLearn, Marker Learning, WireScreen — all of them. One company building geospatial visualization at the frontier of AI was explicit: they listed TypeScript/Next.js/Vercel/Supabase as their stack and said the role requires someone who’s “AI-native in their workflow.”
Python still owns the model layer. But the full-stack application layer — what startups actually ship — is TypeScript. If you’re a Latin American developer still debating JavaScript versus TypeScript, the market has already voted.
4. Remote work is alive, but the wall is real
The good news for Latin American developers: remote work is still the norm for a significant portion of these roles. Aha!, one of the most established companies in the thread, explicitly lists North and South America as their hiring geography. It’s a real opportunity.
The bad news: “REMOTE (US only)” is becoming more frequent, and the wall is getting harder. Companies working with compliance, data residency, or government contracts are tightening geographic requirements. The practical implication: if you’re a Latin American developer targeting the HN pool, filter for roles that explicitly say “North or South America” or simply “REMOTE” without the US restriction. They exist — but you have to look for them.
5. Fluency in AI-native workflows is already a hiring signal
This one surprised me. Schema Labs — a civic tech nonprofit building an AI platform for Greek municipalities — mentioned in their March posting that they use Claude Code and Cursor daily. It wasn’t in the job requirements; it was in the company description. Signal: they want to hire people who see that as a positive.
One of the March postings for an AI geospatial full-stack role was equally explicit: candidates must be “AI-native in their workflow.” Not “comfortable with AI tools.” Native. Companies that reorganized their engineering process around AI-assisted development are already using it as a cultural filter — they don’t want to onboard someone who’s going to resist it.
If you’re not yet using Claude Code, Cursor, or a comparable tool as part of your daily development loop, that gap is becoming visible to the companies that matter.
What to watch in the next 48 hours
This article is a snapshot. The April thread just kicked off this morning and will accumulate hundreds of comments over the next two days. I’ll be back with a follow-up once the thread matures — with more data, more patterns, and a fuller read of what the market is saying about Q2 2026.
The HN thread is public and free: