What's the best programming language for coding agents?

📝 Discussion Summary (Click to expand)

1. Standardized language benchmarking

“equivalent accomplishments in each language… I think that would add in some ~normalization.” – summarybot

2. User‑stylesheet debate & design preferences

“Supply my own stylesheet? No thank you, I'm not here to do work for free.” – _doctor_love

3. Token‑efficiency and language suitability for LLMs

“omitting explicit type declarations makes the code more compact.” – aleph_minus_one

🚀 Project Ideas

A command‑line / web tool that normalizes and compares token usage of code snippets across programming languages (C, Rust, Python, etc.).
Core value: Quantifies “token cost” for LLM‑generated code so developers can choose the most efficient language for AI‑heavy tasks.

Key	Value
Target Audience	LLM developers, AI researchers, language designers
Core Feature	Takes a short snippet (e.g., a memoized Fibonacci implementation) and outputs token count, per‑character density, and normalized score
Tech Stack	Node.js backend, React UI, Rust tokenizer library, SQLite for storing benchmark results
Difficulty	Medium
Monetization	Revenue-ready: Subscription (monthly/annual tiers for private benchmark reports)

HN users repeatedly asked for “equivalent accomplishments in each language” and highlighted token‑efficiency concerns; TokenMetrics directly answers that request.
The tool would let users plug in their own snippets and instantly see which language yields the smallest token footprint, satisfying the demand for a “benchmark” as mentioned by quinnjh and aleph_minus_one.

A browser extension (Chrome/Firefox) that automatically adds a minimal, responsive CSS margin and font‑size rules to any article‑style webpage, making dense layouts like Dan Luu’s blog far more readable.
Core value: Turns “raw, unprepared ingredients” into a pleasant reading experience without requiring users to write their own stylesheet.

Key	Value
Target Audience	General web readers, especially those who frequent long‑form technical blogs
Core Feature	Injects a 5% margin, optimal line‑height, and optional dark‑mode toggle; can be toggled on/off with a single click
Tech Stack	Manifest V3 extension, plain CSS injected via content script, optional user‑style marketplace
Difficulty	Low
Monetization	Hobby

Multiple HN commenters (e.g., lemming, scared_together) expressed frustration with dense layouts and the need to “write a stylesheet for free.” StyleInject solves this by automating the fix.
The extension could also host a community‑curated collection of “best‑looking” styles, appealing to the desire for a simple, one‑click solution.

A SaaS platform that recommends the most LLM‑friendly language for a given task, provides starter project templates, and includes built‑in syntax validation & test harnesses.
Core value: Reduces trial‑and‑error by surfacing languages like Go, Rust, or Gleam that have strong tooling for AI‑generated code, based on community data.

Key	Value
Target Audience	AI‑assisted developers, indie hackers, small dev teams
Core Feature	Input: task description; Output: language recommendation, boilerplate repo, CI pipeline that fails on compile/type errors, optional pricing‑tier for private repos
Tech Stack	FastAPI backend, Dockerized CI with Rust, Go, and TypeScript runtimes, React dashboard
Difficulty	High
Monetization	Revenue-ready: Subscription (team plan with private templates and analytics)

The discussion stressed that “static typed, compiled languages” are better for LLMs to avoid silent failures (serf, frollogaston). This service codifies that insight, giving users a “recommended language” based on empirical token‑efficiency and tooling quality.