How many Science, Security & Hardware use cases are there?

There are 6 use cases in this category right now. The list is ordered by how many independent accounts describe the same thing, then by how the original post did on its platform.

Which agents show up most for this kind of work?

The named agents used most often here are Claude Code, Codex, Claude. Other entries use a custom-built or unnamed agent.

Which apps do these agents touch?

The apps that come up most are GitHub, Discord. Every entry lists the apps the agent actually used, so you can see what you would need to connect.

How does a use case get onto this page?

It has to be something a person reported doing. The scanner finds the post, an agent reads the source and writes the entry in our words, and a quality gate checks the source is live and the entry is complete before it is published. 3 of these are described in more than one independent account.

No. These are reports of what someone else did with an agent. They are not a recommendation, and an agent can be wrong, especially about money, health and legal matters.

AI agents for Science, Security & Hardware: 6 real use cases

Vulnerability hunting, reverse engineering, maths, CAD and robots. Most used named agents: Claude Code, Codex, Claude.

All Science, Security & Hardware use cases

Find an exoplanet candidate in TESS data and win telescope timeNew

A Redditor used Claude Code to comb NASA TESS light curves, found a 3.18-day transit signal on a star 116 light-years away across three years of data, and got an approved TESS observing proposal to test it.

Claude Code

Codex

Reddit · Oct 2026 · 2 accounts

Decompile a first-person shooter to byte-exact C++ with a fleet of agentsNew

Maurice Heumann ran up to 16 Claude Code and Codex agents for three months, coordinated over GitHub issues and Discord, and reports 99% of an old shooter's functions reconstructed, 83% of them byte-exact.

Claude Code

Codexin

GitHub

Discord

momo5502.com · Oct 2026 · 2 accounts

Screen crystals for room-temperature magnetic semiconductors with DFTNew

Geby Jaff at Vals AI ran a team of Claude Opus 5.5 agents through 868 DFT simulations on cloud CPUs; they designed one new compensated magnet and flagged a 1999 Prussian-blue analogue as a candidate.

Claude

vals.ai · Oct 2026 · 2 accounts

Generate buildable Lego CAD models from a text promptNew

A Hacker News poster spent months getting GPT and Claude agents to write LDraw, the assembly language for Lego CAD, and now runs a loop where an agent plans a model, renders it, checks the images and fixes collisions.

Custom agent

Hacker News · Oct 2026

Write firmware for an e-ink fridge-magnet shopping listNew

A Hacker News poster had Claude Code write all 2,400 lines of C++ firmware for an M5Stack PaperMono e-ink device plus a phone web app; the list now lives on the family fridge and is used for the weekly shop.

Claude Code

Hacker News · Sep 2026

Put an AI agent on a Nokia 110 to run phone functionsNew

anupray reverse-engineered the firmware of a 48 MB Nokia 110 4G and loaded a tiny custom agent that sends typed questions to DeepSeek and runs tool calls for the torch, alarm, battery and calls.

Custom agent

Hacker News · Oct 2026

Questions

How many Science, Security & Hardware use cases are there?: There are 6 use cases in this category right now. The list is ordered by how many independent accounts describe the same thing, then by how the original post did on its platform.
Which agents show up most for this kind of work?: The named agents used most often here are Claude Code, Codex, Claude. Other entries use a custom-built or unnamed agent.
Which apps do these agents touch?: The apps that come up most are GitHub, Discord. Every entry lists the apps the agent actually used, so you can see what you would need to connect.
How does a use case get onto this page?: It has to be something a person reported doing. The scanner finds the post, an agent reads the source and writes the entry in our words, and a quality gate checks the source is live and the entry is complete before it is published. 3 of these are described in more than one independent account.
Is this advice?: No. These are reports of what someone else did with an agent. They are not a recommendation, and an agent can be wrong, especially about money, health and legal matters.