35 Developer Tools From Hacker News That Actually

Github Awesome just dropped a rapid-fire showcase of 35 projects pulled straight from Hacker News's Show HN section, and it's honestly one of the better snapshots of where developer tooling is right now. Not vaporware. Not concepts. Actual working code people are shipping.

The format is relentless—14 minutes covering 35 projects means you get about 25 seconds per tool. That sounds chaotic, but it works because these aren't products that need deep dives. They're solutions to specific, annoying problems that developers recognize immediately.

The AI Agent Tooling Explosion

What's striking is how many of these projects exist specifically because AI coding agents are now a real part of people's workflows. Not a future thing—a right now thing that needs infrastructure.

Git Surgeon is the clearest example. It's a set of git primitives designed for autonomous coding agents to organize their commits cleanly. The problem it solves is embarrassingly specific: AI agents write functional code but commit it like chaos gremlins. As the narrator puts it, Git Surgeon gives agents "the tools to slice, amend, and organize commits cleanly as they work. The result is a repository history that doesn't look like a bot was in charge, even when one was."

That's not theoretical. That's someone who's been using Claude or Cursor for real work and got tired of the mess.

Agent Kernel takes a different angle on the same problem space—persistent memory. The video describes it as "three markdown files and a git repo. Clone it, point clawed code or cursor at it, and the agent figures out how to use it on its own." No vector database. No weekend of setup. Just markdown files that maintain identity and track world state.

FlowState packages that same idea as an MCP server so your agent "retains memory about your project architecture, your coding preferences, and outstanding bugs across sessions." Pick up Monday where you left off Friday without re-explaining the entire codebase.

The pattern here is clear: developers are building infrastructure for AI agents the same way they built infrastructure for microservices five years ago. Not because it's trendy, but because the pain points are real enough that spending a weekend building a tool is worth it.

When Jokes Are Also Good Engineering

Crack deserves special mention because it's completely useless and also kind of brilliant. It taps into the undocumented lid angle sensor on MacBooks and maps the opening angle to a creaking door sound in real time. "Open the lid slowly, get a long, horrifying squeak, slam it shut, get the full effect. Completely useless, technically impressive, guaranteed to clear out a coffee shop."

The video calls it "one of the better ways anyone has spent a weekend," and honestly, yeah. It's the kind of project that demonstrates actual technical chops—reverse engineering an undocumented sensor, real-time audio mapping—while having zero practical value. Those projects matter because they're pure creative problem-solving without the constraints of utility.

Jensenify-MCP is in the same category but pushes it further. It injects 2.9 million tokens of Homer, Shakespeare, Tolstoy, and the King James Bible into every AI tool call "so your coding agent has deep humanistic grounding when it fixes a null pointer exception." The creator acknowledges it costs a fortune in API tokens. That's the joke and also entirely the point.

The Stuff That Just Quietly Works

Not everything needs to be clever. Some of the most useful tools in this list are the ones solving boring problems really well.

Meow SSH is a terminal UI for managing SSH connections that's "fast, organized, and wrapped in a cat themed aesthetic that makes it considerably more enjoyable than it has any right to be." Your SSH config file is probably a graveyard of IP addresses you don't remember and keys you can't locate. This fixes that. Not revolutionary, just actually useful.

DockAutoHide watches your screen and toggles macOS dock autohide only when a window overlaps it. "The kind of feature that feels obvious in hindsight and should have shipped with Mac OS years ago." It probably should have.

ThermalMarky turns your abandoned thermal printer into something useful by accepting markdown and spitting out formatted receipts. "Everyone has a cheap thermal printer sitting in a drawer after two uses," the narrator notes. If you're one of those people, this is your excuse to dig it back out.

The Actually Ambitious Stuff

Then there's the projects that are genuinely pushing technical boundaries.

Threadprocs is an experimental Linux project that lets multiple independent executables share a single virtual address space. The practical result: "you can pass raw C++ or Rust pointers between completely separate programs with zero copy overhead. No serialization, no shared memory gymnastics, no IPC overhead, just a pointer passed directly."

That's not a incremental improvement. That's rethinking how process isolation works at a fundamental level.

Refrax is similarly ambitious—a new browser that uses an undocumented private macOS class called KPortalLayer to mirror GPU memory of a single web page across multiple windows. "Hundreds of live updating tabs, RAM footprint of one." Whether that's actually sustainable or just a clever hack that'll break with the next macOS update is an open question, but the engineering is genuinely impressive.

M33MU is a full ARM Cortex-M33 emulator written in C with "TrustZone awareness built-in, GDB remote server, terminal UI and execution recording for reverse stepping through firmware bugs." For firmware developers debugging embedded systems without physical hardware, that's not just useful—it's potentially workflow-changing.

What This Collection Actually Shows

Looking at all 35 projects together, a few patterns emerge. First, the barrier to shipping useful tools is lower than it's ever been. These aren't venture-backed startups with years of runway. They're weekend projects that solve specific problems well enough to share.

Second, AI coding tools have created an entirely new category of infrastructure needs. Memory management, context persistence, commit organization, prompt optimization—none of these were problems two years ago. Now they're annoying enough that multiple people are building competing solutions.

Third, the line between "useful" and "just for fun" is increasingly blurry, and that's fine. Crack serves no purpose. Jensenify-MCP is an expensive joke. But they're both technically interesting and they make the ecosystem more enjoyable to be part of.

The Show HN section has always been a decent temperature check on what developers are actually building versus what tech media thinks they're building. This particular batch suggests the boring answer: people are solving the annoying problems right in front of them, building infrastructure for AI agents they're already using, and occasionally having fun with technically impressive nonsense.

Which is probably healthier than whatever the alternative would look like.

— Tyler Nakamura, Consumer Tech & Gadgets Correspondent