What Indie Hackers Are Building in March 2026

What’s Happening

Every few months, the Hacker News community surfaces one of the internet’s most honest barometers of startup activity: the recurring “Ask HN: What Are You Working On?” thread. The March 2026 edition just dropped, and it’s already generating hundreds of responses from solo developers, early-stage founders, and side-project tinkerers sharing exactly what they’re building — and why. If you want to know what the early-adopter builder class cares about before it becomes a TechCrunch headline, this is where you look.

What’s immediately striking about this month’s thread is the degree to which AI-native products have moved from novelty to infrastructure. A year ago, half the responses were “I’m building a ChatGPT wrapper.” That era is emphatically over. In March 2026, builders are applying large language models and agentic frameworks to deeply vertical problems — compliance workflows for mid-market finance teams, automated QA pipelines for hardware manufacturers, AI copilots for niche professional workflows that most VCs wouldn’t even think to look at. The wrappers are dead; the picks-and-shovels phase is fully underway.

Equally notable: a visible surge in local-first software, privacy-centric tooling, and products explicitly designed to work without cloud dependency. The post-GDPR, post-AI-training-data-panic world has clearly shifted builder sentiment. There’s also a thread running through dozens of posts about fatigue with subscription pricing — founders are building tools with one-time purchases, lifetime deals, and usage-based models as a deliberate competitive wedge.

Why Founders Should Care

These threads aren’t just community navel-gazing. They function as a crowdsourced early warning system. Projects mentioned in past “What Are You Working On?” threads have gone on to become YC-backed companies, ProductHunt sensations, and acquisition targets. If you’re a founder trying to read the market, this is primary-source intelligence — hundreds of smart, technical builders telling you, unprompted, where they think opportunity lives.

The timing matters, too. According to PitchBook’s Q1 2026 Emerging Tech Report, pre-seed and seed deal volume has ticked up 14% year-over-year, with the sharpest growth in vertical AI SaaS and developer infrastructure. What builders post in these threads in March often maps to what raises a round by September. The signal-to-noise ratio is unusually high because most respondents aren’t pitching — they’re just sharing work in progress, which means you get a more honest picture of conviction.

The market size conversation has shifted as well. We’re past the era where every AI startup needed to claim a $50 billion TAM to get a meeting. Investors — especially at the seed stage — are increasingly comfortable with smaller, well-defined markets that AI tooling can dominate efficiently. A compliance automation tool for dental practices? A local-first CRM for independent insurance agents? These ideas show up repeatedly in the March thread, and they map directly to the kind of focused, capital-efficient bets that are actually getting funded right now.

Who’s Already Moving

Several companies and projects that have emerged from this builder ecosystem are worth watching closely:

• Luminai — Originally shared in a 2024 HN thread, Luminai has grown into a serious player in AI-driven process automation for enterprise. They went through YC and recently closed a Series A. Their origin story is a textbook example of how a side project mentioned on HN can become a funded company in under two years.
• Anyquery — A developer tool that lets you query anything (APIs, local files, databases) with SQL. It’s been mentioned multiple times in HN builder threads and has cultivated a loyal open-source community. It represents the developer infrastructure wave perfectly: unsexy, deeply useful, built by one person.
• Hatchet — An open-source task queue and workflow engine designed for AI workloads. As agentic AI systems become more complex, orchestration tooling like Hatchet has quietly become critical infrastructure. They’ve been gaining traction among builders frustrated with existing queue systems that weren’t designed for LLM-scale unpredictability.
• Bolna — A platform for building voice-based AI agents, specifically targeting customer service and sales automation. Voice AI is a category that keeps appearing in these threads, and Bolna is one of the more polished entries. They’re competing in a space that a16z’s latest AI report flagged as one of the fastest-growing application layers.

“The best startups don’t come from brainstorming sessions — they come from someone scratching their own itch and posting about it on a forum at 2 a.m.” — paraphrased from a frequent sentiment in the HN builder community

The Startup Opportunity

If you’re a founder reading this thread closely — and you should be — here are the specific patterns worth acting on:

• Vertical AI agents for regulated industries. Healthcare billing, legal document review, insurance underwriting, tax preparation. The March thread is packed with builders targeting these niches. The opportunity isn’t the AI model itself — it’s the domain expertise wrapped around it. If you’ve spent years in a regulated field and understand the workflow pain, you have a moat that a generalist AI startup simply can’t replicate.
• Local-first and offline-capable tools. There’s a genuine counter-movement forming against cloud-dependent SaaS. Builders are shipping note-taking apps, project management tools, and even CRMs that work entirely on-device. The pitch is simple: your data never leaves your machine. For enterprise buyers spooked by AI training data controversies, this is increasingly compelling.
• One-time-purchase software. Subscription fatigue is real, and several builders in the thread are explicitly positioning against it. Tools like Raycast and Obsidian proved that premium one-time-purchase software can build sustainable businesses. If you can find a category where incumbents charge $20/month for something that should cost $49 once, there’s an opening.
• AI infrastructure for non-AI companies. Most businesses don’t want to hire an ML team. They want a plug-in that makes their existing workflow smarter. Think: AI-powered anomaly detection that drops into an existing Postgres database, or an LLM layer that auto-categorizes support tickets in Zendesk. The picks-and-shovels opportunity in AI middleware is enormous and undercovered.
• Developer experience tooling. Every wave of new technology creates a developer tooling opportunity. With agentic frameworks, new programming paradigms around prompt engineering, and the explosion of multi-modal AI, developers need better debugging, testing, observability, and deployment tools. This is a category where small teams can build outsized value fast.

Watch Out For

Not everything in these threads is signal. Some honest caveats:

Survivorship bias is real. For every Luminai that goes from HN thread to Series A, hundreds of projects die quietly. The enthusiasm in these threads can make it feel like every niche AI tool is a viable business. Many aren’t. The gap between “cool side project” and “something people will pay for” remains as wide as ever.

The AI infrastructure layer is consolidating fast. If you’re building developer tooling for AI workloads, be aware that the major cloud providers — AWS, Google Cloud, Azure — are aggressively absorbing this layer. What’s an indie project today could be a built-in feature of Bedrock or Vertex AI tomorrow. Speed to market and community lock-in are your only real defenses.

Local-first is harder than it looks. Syncing, conflict resolution, and offline-first architectures are genuinely difficult engineering problems. Several builders in the thread acknowledge struggling with these challenges. If you’re going this route, budget for significantly more technical complexity than a standard cloud SaaS app.

Pricing innovation can backfire. One-time purchases sound great for users, but they create a revenue model that requires constant new customer acquisition. Unless you have a strong upgrade path or complementary revenue stream, the economics can get brutal at scale. Don’t mistake “people say they want this” for “this is a sustainable business model.”

Regulatory risk in AI-adjacent spaces. If you’re building AI tools for healthcare, finance, or legal — the hottest verticals in the thread — you need to be watching the regulatory environment closely. The EU AI Act is now being enforced, and U.S. state-level AI regulation is accelerating. Building something that becomes non-compliant six months after launch is a founder’s nightmare.

The Bottom Line

The March 2026 “What Are You Working On?” thread on Hacker News isn’t just community show-and-tell — it’s a real-time map of where technical founders see unmet demand. The macro signal is clear: the gold rush phase of AI is over, and we’re entering the craft phase. Builders are going narrower, deeper, and more opinionated about the problems they solve and how they charge for it. The winners in this cycle won’t be the ones with the most powerful model; they’ll be the ones who understand a specific user’s workflow better than anyone else.

If you’re a founder, here’s what to do this week: read the thread yourself. Don’t skim — read the comments, the reply chains, the debates about pricing and architecture. Then ask yourself: which of these problems do I understand better than the person building the solution? That’s where your edge is. The best startup ideas aren’t invented in a vacuum — they’re spotted in the gap between what a builder is attempting and what the market actually needs. That gap is on full display right now.