frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

Open in hackernews

AetherScript: AI-Assisted Development That You Can Trust

1•papabili•6h ago
AetherScript: AI-Assisted Development That You Can Trust Tired of the "magic" and unpredictability of AI code generation? Do you love the power of LLMs but hate the messy diffs, the loss of control, and the "hallucinated" code that pollutes your pristine codebase?

What if we could collaborate with AI the same way we collaborate with human developers? With clear contracts, explicit reviews, and a process that guarantees human oversight?

This is the vision behind AetherScript.

AetherScript introduces a new, structured workflow for human-AI collaboration, built on a simple yet powerful philosophy: separate human intent from AI implementation.

How It Works: The Acceptance Model You Define the Intent (.as files): In standard TypeScript files (.as), you write the high-level architecture, interfaces, and method signatures. You define the "what." For the parts you want to delegate, you leave a simple placeholder.

// You write this in user.service.as export class UserService { public async create(data: UserCreateInput): Promise<User> { // AI, please implement the user creation logic. return "${this.create(data)}"; } }

AI Fills the Implementation (.asc files): The AetherScript engine generates the corresponding implementation in a separate, companion file (.asc). This file acts as a sandboxed staging area for the AI's work. It's the "how."

You Review and Accept (aesc merge): The AI's code never touches your main source until you explicitly approve it. You can review the implementation in a clean, focused diff view. When you're satisfied, a single command, aesc merge, seamlessly integrates the code, replacing the placeholder. It's like a pull request from your AI partner.

Why AetherScript? Full Control & Trust: No more unpredictable code magically appearing in your files. You are always the final gatekeeper.

Clean Git History: Merges are atomic and meaningful, showing a clear transition from an intent to a completed implementation.

Deep Toolchain Integration: Designed from the ground up with a compiler, a Bun runtime plugin for a seamless dev experience, and a powerful VS Code extension in mind.

A Structured Philosophy: AetherScript isn't just a tool; it's a complete methodology for making AI a true, reliable partner in professional software engineering.

We believe this is the future of AI-assisted development—structured, predictable, and always developer-led.

Intrigued? We've laid out the entire vision and technical architecture in our design document.

Dive deeper and join the conversation on GitHub: https://github.com/Pabears/AetherScript/tree/main We welcome your feedback, ideas, and contributions!

Show HN: Yet another JSON lib (C++17)

https://github.com/DmitriBogdanov/UTL/blob/master/docs/module_json.md
1•GeorgeHaldane•7m ago•0 comments

Are China’s universities really the best in the world?

https://www.economist.com/science-and-technology/2025/06/18/are-chinas-universities-really-the-best-in-the-world
2•rustoo•8m ago•0 comments

How to Use T-SNE Effectively

https://distill.pub/2016/misread-tsne/
1•rsecora•9m ago•0 comments

Far-Right Extremists Use Minecraft to Gamify Radicalisation

https://gnet-research.org/2025/07/02/playing-with-hate-how-far-right-extremists-use-minecraft-to-gamify-radicalisation/
1•Improvement•12m ago•0 comments

Microsoft to Lay Off About 9k Employees

https://www.nytimes.com/2025/07/02/technology/microsoft-layoffs-ai.html
1•reaperducer•15m ago•0 comments

The Dichotomy of Mario Kart World

https://ravi64.com/dichotomy-mario-kart-world/
3•merlioncity•15m ago•0 comments

China's Evolving Industrial Policy for AI

https://www.rand.org/pubs/perspectives/PEA4012-1.html
2•ironyman•20m ago•0 comments

Love and Other Unrealizable Utopian Projects

https://agoodhardstare.substack.com/p/love-and-other-ill-fated-utopian
1•jger15•28m ago•0 comments

Marginalized Americans are highly skeptical of artificial intelligence

https://news.umich.edu/marginalized-americans-are-highly-skeptical-of-artificial-intelligence/
1•Improvement•32m ago•0 comments

Show HN: Recivo – Receive Emails Programmatically

https://recivo.email/
5•WilcoKruijer•44m ago•0 comments

Building Linux kernel on macOS natively

https://seiya.me/blog/building-linux-on-macos-natively
1•ingve•45m ago•0 comments

Bad Review Sniper – AI tool to detect and fix bad reviews automatically

https://cheery-mooncake-0faf8f.netlify.app/
1•Farid_agha87•45m ago•1 comments

One line email taught me more about communication than months of building

2•lebonnnn•46m ago•0 comments

How Oracle Is Winning the AI Compute Market

https://semianalysis.com/2025/06/30/how-oracle-is-winning-the-ai-compute-market/
2•nsoonhui•46m ago•0 comments

The Flaw in the Kelly Criterion – Betting Under Uncertainty

https://www.architect.co/posts/the-flaw-in-the-kelly-criterion
2•auc•46m ago•0 comments

Ask HN: Ideas to acquire "good taste" in programming?

1•danielciocirlan•49m ago•0 comments

How China's new auto giants left GM, VW and Tesla in the dust

https://www.reuters.com/investigations/how-chinas-new-auto-giants-left-gm-vw-tesla-dust-2025-07-03/
4•pseudolus•50m ago•0 comments

How I keep up with AI progress (and why you must too)

https://blog.nilenso.com/blog/2025/06/23/how-i-keep-up-with-ai-progress/
2•ananthrk•55m ago•0 comments

A link is all you need

https://aifoc.us/a-link-is-all-you-need/
1•freediver•58m ago•0 comments

ChatGPT creates phisher's paradise by recommending the wrong URLs for banks

https://www.netcraft.com/blog/large-language-models-are-falling-for-phishing-scams
4•bundie•1h ago•0 comments

Newly Discovered 'Hyperaccessible' DNA State May Revolutionize Cancer Treatment

https://www.sciencedaily.com/releases/2025/01/250121162043.htm
1•karlperera•1h ago•1 comments

An optimization and debugging story with Go and DTrace

https://gaultier.github.io/blog/an_optimization_and_debugging_story_go_dtrace.html
2•broken_broken_•1h ago•0 comments

Pure CSS Moon Phases

https://codepen.io/xaelan/pen/bjqOvo
4•ludicrousdispla•1h ago•0 comments

On July 7, Gemini AI will access WhatsApp and more. Learn to disable on Android

https://tuta.com/blog/how-to-disable-gemini-on-android
18•shaunpud•1h ago•1 comments

Show HN: GPT Watermark-Remover React/Next.js tool to strip hidden AI watermarks

https://gpt-watermark-remover.com/
1•AleksDoesCode•1h ago•0 comments

Show HN: Soham Parekh AI

https://www.thesohamparekh.com/
4•the2ndfloorguy•1h ago•0 comments

Ksymless – A kernel rootkit works without kallsyms

https://github.com/rota1001/ksymless
1•rota1001•1h ago•0 comments

Workers training Meta's AI in Ireland speak out

https://www.thejournal.ie/meta-workers-ireland-6745653-Jul2025/
3•austinallegro•1h ago•0 comments

A foundation model to predict and capture human cognition

https://www.nature.com/articles/s41586-025-09215-4
1•crocowhile•1h ago•1 comments

Dutch authorities: Almost every Dutch citizen has too much PFAS in their blood

https://nos.nl/artikel/2573446-rivm-bijna-iedereen-in-nederland-heeft-te-veel-pfas-in-bloed
3•elisaado•1h ago•1 comments