frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

An open replacement for the IBM 3174 Establishment Controller

https://github.com/lowobservable/oec
1•bri3d•1m ago•0 comments

The P in PGP isn't for pain: encrypting emails in the browser

https://ckardaris.github.io/blog/2026/02/07/encrypted-email.html
1•ckardaris•3m ago•0 comments

Show HN: Mirror Parliament where users vote on top of politicians and draft laws

https://github.com/fokdelafons/lustra
1•fokdelafons•4m ago•1 comments

Ask HN: Opus 4.6 ignoring instructions, how to use 4.5 in Claude Code instead?

1•Chance-Device•5m ago•0 comments

We Mourn Our Craft

https://nolanlawson.com/2026/02/07/we-mourn-our-craft/
1•ColinWright•8m ago•0 comments

Jim Fan calls pixels the ultimate motor controller

https://robotsandstartups.substack.com/p/humanoids-platform-urdf-kitchen-nvidias
1•robotlaunch•11m ago•0 comments

Exploring a Modern SMTPE 2110 Broadcast Truck with My Dad

https://www.jeffgeerling.com/blog/2026/exploring-a-modern-smpte-2110-broadcast-truck-with-my-dad/
1•HotGarbage•11m ago•0 comments

AI UX Playground: Real-world examples of AI interaction design

https://www.aiuxplayground.com/
1•javiercr•12m ago•0 comments

The Field Guide to Design Futures

https://designfutures.guide/
1•andyjohnson0•13m ago•0 comments

The Other Leverage in Software and AI

https://tomtunguz.com/the-other-leverage-in-software-and-ai/
1•gmays•15m ago•0 comments

AUR malware scanner written in Rust

https://github.com/Sohimaster/traur
3•sohimaster•17m ago•1 comments

Free FFmpeg API [video]

https://www.youtube.com/watch?v=6RAuSVa4MLI
3•harshalone•17m ago•1 comments

Are AI agents ready for the workplace? A new benchmark raises doubts

https://techcrunch.com/2026/01/22/are-ai-agents-ready-for-the-workplace-a-new-benchmark-raises-do...
2•PaulHoule•22m ago•0 comments

Show HN: AI Watermark and Stego Scanner

https://ulrischa.github.io/AIWatermarkDetector/
1•ulrischa•22m ago•0 comments

Clarity vs. complexity: the invisible work of subtraction

https://www.alexscamp.com/p/clarity-vs-complexity-the-invisible
1•dovhyi•23m ago•0 comments

Solid-State Freezer Needs No Refrigerants

https://spectrum.ieee.org/subzero-elastocaloric-cooling
2•Brajeshwar•24m ago•0 comments

Ask HN: Will LLMs/AI Decrease Human Intelligence and Make Expertise a Commodity?

1•mc-0•25m ago•1 comments

From Zero to Hero: A Brief Introduction to Spring Boot

https://jcob-sikorski.github.io/me/writing/from-zero-to-hello-world-spring-boot
1•jcob_sikorski•25m ago•1 comments

NSA detected phone call between foreign intelligence and person close to Trump

https://www.theguardian.com/us-news/2026/feb/07/nsa-foreign-intelligence-trump-whistleblower
9•c420•26m ago•1 comments

How to Fake a Robotics Result

https://itcanthink.substack.com/p/how-to-fake-a-robotics-result
1•ai_critic•26m ago•0 comments

It's time for the world to boycott the US

https://www.aljazeera.com/opinions/2026/2/5/its-time-for-the-world-to-boycott-the-us
3•HotGarbage•27m ago•0 comments

Show HN: Semantic Search for terminal commands in the Browser (No Back end)

https://jslambda.github.io/tldr-vsearch/
1•jslambda•27m ago•1 comments

The AI CEO Experiment

https://yukicapital.com/blog/the-ai-ceo-experiment/
2•romainsimon•28m ago•0 comments

Speed up responses with fast mode

https://code.claude.com/docs/en/fast-mode
5•surprisetalk•32m ago•1 comments

MS-DOS game copy protection and cracks

https://www.dosdays.co.uk/topics/game_cracks.php
4•TheCraiggers•33m ago•0 comments

Updates on GNU/Hurd progress [video]

https://fosdem.org/2026/schedule/event/7FZXHF-updates_on_gnuhurd_progress_rump_drivers_64bit_smp_...
2•birdculture•34m ago•0 comments

Epstein took a photo of his 2015 dinner with Zuckerberg and Musk

https://xcancel.com/search?f=tweets&q=davenewworld_2%2Fstatus%2F2020128223850316274
14•doener•34m ago•2 comments

MyFlames: View MySQL execution plans as interactive FlameGraphs and BarCharts

https://github.com/vgrippa/myflames
1•tanelpoder•35m ago•0 comments

Show HN: LLM of Babel

https://clairefro.github.io/llm-of-babel/
1•marjipan200•35m ago•0 comments

A modern iperf3 alternative with a live TUI, multi-client server, QUIC support

https://github.com/lance0/xfr
3•tanelpoder•37m ago•0 comments
Open in hackernews

Build Software. Build Users

https://dima.day/blog/build-software-build-users/
71•dinerville•1mo ago

Comments

crabmusket•1mo ago
This is interesting, and I think worth trying. However,

    The process is iterative:

    Vibe code users <--> Vibe code software

    Step by step, you get closer to truly understanding your users
Do not fool yourself. This is not "truly" "understanding" your "users". This is a model which may be very useful, but should not be mistaken for your users themselves.

Nothing beats feedback from humans, and there's no way around the painstaking effort of customer development to understand how to satisfy their needs using software.

bahmboo•1mo ago
I agree. I do like the general idea as an exploration.

Perhaps the idea is to use an LLM to emulate users such that some user-based problems can be detected early.

It is very frustrating to ship a product and have a product show stopper right out of the gate that was missed by everyone on the team. It is also sometimes difficult to get accurate feedback from an early user group.

fainpul•1mo ago
This is ridiculous. I doubt this would work with a general AI, but it surely cannot work with LLMs who understand exactly nothing about human behaviour.
josters•1mo ago
They may not understand it but they may very well be able to reproduce aspects of feedback and comments on similar pieces of software.

I agree that the approach shouldn’t be done unsupervised but I can imagine it being useful to gain valuable insights for improving the product before real users even interact with it.

fainpul•1mo ago
> reproduce aspects of feedback and comments on similar pieces of software

But this is completely worthless or even misleading. There is zero value in this kind of "feedback". It will produce nonsense which sounds believable. You need to talk to real users of your software.

ZachSaucier•1mo ago
Letting the LLM generate use cases is probably not a good idea. But writing common use cases and having a bot crawl your app and then write up how many steps it took to accomplish the goal is a good idea that I hadn't thought of before. You could even set it up as a CI check to make sure that new features that introduce more steps for specific flows are very conscious decisions. In a large application this could be a very useful feature.
wiseowise•1mo ago
Another fart in the wind. How to write lots of “programming philosophy” and say nothing.
invide•1mo ago
Good point. I think it is the time to remove the line between the engineering and product management completely. Because, we can.
elxr•1mo ago
A bit too vague to be useful advice don't you think?

Why not show some actual examples of these agents actually doing what you describe. How exactly would you set up an agent to simulate a user?

crabmusket•1mo ago
To me it sounds like one way to do this would be to have LLMs write Cucumber test cases. Those are high level, natural language tests which could be run in a browser.
Sharanxxxx•1mo ago
Yeah, I have build a Product hunt alternative for solo founders - Solo Launches to give them visibility. I built 290+ users from it till now and its free and giving a good DR dofollow backlink. https://sololaunches.com
oldsj•1mo ago
I like the idea. As a solo dev I've experimented with creating Claude subagents for multiple perspectives for "team leads" and will run ideas through them (in parallel). The subagents are just simple markdown explaining the various perspectives that are usually in contention when designing stuff. And a 'decider' that gives me an executive summary.

  agents/
    |-- customer-expert.md - validates problem assumptions, customer reality
    |-- design-lead.md - shapes solution concepts, ensures UX quality
    |-- growth-expert.md - competitive landscape, positioning, distribution
    |-- technical-expert.md - assesses feasibility, identifies technical risks
    |-- decider-advisor.md - synthesizes perspectives, executive analysis
bisonbear•1mo ago
I've experimented with something similar - my flow is to have the subagents "initialize" a persona for the task at hand, and then have the main thread simulate a debate between the personas. Not sure if it's the best approach but it's helpful to get a diversity of perspectives on an issue
nprateem•1mo ago
s/validates/guesses/

Ftfy. You might as well toss a coin.

oldsj•1mo ago
I think we’re slightly better than random at this point
nprateem•1mo ago
You really think asking AI to spaff out some bullshit validates anything?
3D39739091•1mo ago
> LLMs likely have a much better understanding of what our users need and want.

They don't.

Basically this sounds like Agentic Fuzz Testing. Could it be useful? Sure. Does it have anything to do with what real users need or want? Nope.

canadiantim•1mo ago
Pretty radical idea, love it! Definitely going to give this a try