gunta

9 miesięcy temudiscuss

Author here. I analyzed the infrastructure shift happening as AI agents become the majority of web traffic (51% per Imperva 2025).

Key findings: - Token costs drop 95% when serving JSON/Markdown vs HTML - Companies with agent-friendly APIs seeing 400% traffic increases - New metrics (AIX, TX, MPX) becoming more important than traditional KPIs

The article includes: - OpenAPI, Arazzo, MCP and GEPA - Examples for content negotiation - 30-day transformation roadmap

Happy to answer questions about implementation details or the economic implications.

gunta

11 miesięcy temudiscuss

Wait, are you saying you want to help build this? Because YES - I've validated demand (money's already flowing in from the Stripe Fund buttons on the page at the bottom), and I need smart people who get it.

What's your angle on solving this?

gunta

11 miesięcy temudiscuss

Ha! Fair play on the reality check ;)

Though funny thing - while writing that admittedly over-caffeinated post, I was watching something more interesting happen: $400+ in binding commitments rolled in over 24 hours for a GitHub bot that doesn't exist yet. Not "I'm interested" clicks. Actual credit cards, 7 random internet user credit cards.

The prototype that revealed the gap? It wasn't the bot. It was discovering people will pay for software that literally doesn't exist if you just ask them to commit first. The friction isn't in building - it's in validation.

I spent years maintaining a 13k-star project, burning out while users tried to throw money at features. Built this "commitment engine" in 3 hours. Now I'm thinking: what if every open source project could pre-sell features before writing them? What if every startup could get paid before building?

You're right to call out the hype. But sometimes the real billion-dollar opportunity isn't what you're selling - it's what you accidentally discovered while selling it. This validation mechanism might be more valuable than the original product.

"Build nothing → Get paid → Build exactly that" beats "Build everything → Hope someone pays" every time.

But yeah, I should probably calm down with the "every gap is billions" talk. Maybe just millions.

gunta

11 miesięcy temudiscuss

Ha, you know what? Your actually making my case for me. 'Markdown is the universal language of AI' - yeah ok, I got carried away with the marketing speak there. Guilty. But here's the thing tho... every single AI model is literally trained on github. When you ask ChatGPT or Claude to write docs, what format does it use? Markdown. When you feed it code documentation, whats it expecting? Markdown. Its not some cosmic truth, its just... practical reality. But yeah, calling it "universal language" was cringe, I'll give you that. You might remember the recent debate of JSON vs XML vs Markdown: https://x.com/_Anshuman_Jha/status/1953826030084665804 Anyways.

'Vaporware',, This is EXACTLY the problme I'm trying to solve here https://gunta.github.io/PRiority/ !! You know what vaporware actually is? Its when someone builds for 6 months in secret, launches with huge promises, then disappears with peoples money.

What I'm doing is literally the opposite - I'm asking 'will you pay for this?' BEFORE I build anything. If people say no (with their wallets), I build nothing. Zero vaporware created.

Think about it... I've collected hundreds of dollars in 24hrs for something that I explicitly tell people DOESN'T EXIST YET. Thats not vaporware - thats just pre-orders with honesty.

The difference? Traditional vaporware: Build promises → Take money → Never deliver My approach: Take commitments → Hit threshold or refund → Build only what's funded

'Buzzwords and hype' - You got me there. But you know what else is hype? Every successful product launch ever. The difference is whether you deliver after the hype or not.

The github bot? I could literally build it tonight if I wanted. Its not that complicated - parse PRs, check payment status, merge or don't merge. The tech is trivial. The innovation (if you can even call it that) is just asking 'who wants this enough to pay?' before writing a single line of code.

Your skepticism is totally valid btw. We've all seen too much BS in tech. But engagment is signal too - even you cared enough to comment. And honestly? I'd rather have skeptics keeping me honest than cheerleaders enabling vaporware.

The real test: if I don't deliver after taking money, come back and roast me. Screenshot this. I'll deserve it.

gunta

11 miesięcy temudiscuss

You raise a great point - and honestly, it's exactly the reaction that validates why this exists.

You're right that the visible part is just a landing page. But that's like saying Stripe is just a payment form, or that Product Hunt is just a list. The innovation isn't the frontend - it's the system underneath.

What I actually built in 3 hours is a "commitment engine". Not a survey asking 'would you use this?' (everyone says yes), but a system that asks 'will you put money behind this?' The difference between intention and commitment is everything in product validation.

Here's what's actually happening: - In 24 hours, I've collected hundreds of dollars in binding commitments for a product that doesn't exist yet - Each payment is a data point with skin in the game - real signal, zero noise - I got real feedback from real users and I am changing the business model and details of the specs of the Github bot as we talk - The GitHub bot implementation? That's the easy part. I can do that in less than 3 hours. The hard part was proving people will pay for it BEFORE building it

Think about it: How many products die after months of development because nobody actually wanted them? How many 'validate with landing pages' experiments get false positives from people who click 'interested' but never convert?

This flips the entire model. Build nothing → Get paid → Build only what's already sold.

I believe this is the future of how products get built. Not 'build it and they will come,' but 'they came, they paid, now build it.'

so, what's the twist? Once this works, the validation platform itself (product #2) becomes the real business. This would be something like Kickstarter meets AngelList meets GitHub, every idea gets validated with real money before a single line of code is written.

gunta

11 miesięcy temudiscuss

I've analyzed 100 installation methods across all platforms and created a trust scoring system. Key findings: - curl|sh: Trust Score 18 (critically unsafe) - iOS App Store: Trust Score 98 (near perfect) - 80-point gap between best/worst methods on same platform

  As an engineer writing my first paper, I'm looking for:
  1. Security researchers for technical review
  2. Academic co-authors familiar with supply chain security
  3. ArXiv endorsement (cs.CR or similar categories)

  The framework addresses recent incidents (XZ backdoor, CrowdStrike) and helps teams choose secure installation methods. Would love feedback from the community!

gunta

2 lata temudiscuss

Using it in production

gunta

7 lat temudiscuss

There is OpenSTF which is similar but for physically remote controlling Smartphones. No need to install any client. Can be controlled from any browser, even from another smartphone.

https://openstf.io

gunta

7 lat temudiscuss

it's sad to see these "moved from X language to Y language and it became faster!" every-time.

If the CPU is the hugest bottleneck, the best answer is not to optimize the algorithm by going to a lower level language, but rather to invest in a different architecture like GPGPU or FPGA.

For example, this paper shows a significant speed-up for PIP( Polygon in Point) algorithm, going from 15hs (CPU) to a mere 11sec (GPU) in task load-time. https://pdfs.semanticscholar.org/1e51/e3c681e1afc908a41ac253...

gunta

7 lat temudiscuss

Perhaps form a global NPO where participant countries give x% of all their taxes to donate to OSS projects

gunta

7 lat temudiscuss

Ive been thinking that this is really the current best solution to the problem.

I want to try this, is there any project who has implemented this MIT/Apache licenses on request?

Ideally we need first a system to do the request approval a la https://cla-assistant.io

Submissions

51% of web traffic is now AI, but most APIs still return HTML

The Great Inversion: Why Everything You Know About Software Is Now Wrong

Show HN: InstallTrust Score – Quantifying software installation security [pdf]

[untitled]

Architecting Istio 1.1 for Performance

Announcing Istio 1.1

comments