[HN Gopher] Streaming AI agent desktops with gaming protocols
       ___________________________________________________________________
        
       Streaming AI agent desktops with gaming protocols
        
       Author : quesobob
       Score  : 70 points
       Date   : 2025-11-05 16:59 UTC (10 days ago)
        
 (HTM) web link (blog.helix.ml)
 (TXT) w3m dump (blog.helix.ml)
        
       | asmor wrote:
       | > because we're all going to become managers of coding agents
       | whether we like it or not
       | 
       | I will join the woodworking people before that happens, thanks.
        
         | DrewADesign wrote:
         | A career change that left me as a recent graduate in a
         | decimated marketplace missing the bottom ten rungs on the
         | ladder and no interest in getting back into the software world
         | has led me to advanced manufacturing as a metal worker. I code
         | a little, move heavy steel pieces periodically which is a nice
         | way to break up the standing/sitting but not nearly as much as
         | a general laborer, solve lots of problems, keep my trigonometry
         | muscles toned, am forced to take breaks, get paid for my
         | overtime, there's a union that the company ownership is totally
         | willing to work with, and when I'm not at work, work isn't with
         | me. There's something very satisfying about leaving work with
         | exercised muscles, smelling slightly of cutting oil. The money
         | sucks comparatively so early in my career, but the rate
         | increases more for performance than seniority so its rising
         | quickly, the benefits are good, the career trajectory is
         | pointing upwards, and longevity-wise, it's certainly a whole
         | lot better than gig work.
         | 
         | There's a huge crisis in US manufacturing: we're bleeding craft
         | knowledge because off-shoring let companies hire existing
         | experienced workers for decades, so they never had to train a
         | new generation of tradespeople. Now all those folks are dying
         | and retiring and they need people to pick up that deep
         | knowledge quickly. Codifying and automating is going to kill
         | jobs either way, but one factory employing a few people making
         | things for other factories with local materials is better than
         | everything perpetually shifting to the cheap labor market du
         | jour. I'm feeling much more optimistic about the future of this
         | than the future of tech careers.
         | 
         | I think over the next few years, a very large percentage of
         | folks in tech will find themselves on the other side of the
         | fence, quickly realize that their existing expertise doesn't
         | qualify them for any other white collar jobs where vibe coding
         | experience is a bullet point in the pluses section, that tech
         | consulting is declining even faster than salaried jobs, and
         | that they're vastly _less_ qualified than the competition for
         | blue collar jobs. Gonna be a rough road for a lot of folks. I
         | wouldn't invest in SF real estate any time soon.
        
           | seemaze wrote:
           | There's 1,000 established industries that don't offer the
           | rapid growth and pay outs of the modern tech ecosystem. I'm
           | excited to see some of the current industrial backwaters soak
           | up technical talent freed up by the SV AI brain drain.
           | 
           | To think we've handsomely paid our best and brightest the
           | last few decades in pursuit of.. advertising?
        
             | DrewADesign wrote:
             | Hopefully the exodus from the tech industry won't kill
             | demand for too many job markets that are close comfortable
             | cousins to the tech world.
        
             | majormajor wrote:
             | > To think we've handsomely paid our best and brightest the
             | last few decades in pursuit of.. advertising?
             | 
             | I think "efficiency" is more accurate there. Even post-
             | Google/ad-tech-boom the overall trends that started decades
             | earlier continued to be: (1) faster turnaround time on
             | communications, (2) faster delivery of result artifacts,
             | (3) faster knowledge of changes in the market and faster
             | response.
             | 
             | Advertising is a particularly visible field with lots of
             | money to throw at those things (active investment trading
             | is another). But practically every other industry has
             | chased those same things as well, all the way down to
             | things like parking meters.
             | 
             | Personally I'm not convinced that this is such a great
             | thing _anyway_ - does anyone _enjoy_ their boss messaging
             | you at 11PM on any day they want whenever they get the
             | fancy? - but that 's the larger reason so much brainpower
             | has been invested into it.
        
             | tomnipotent wrote:
             | > don't offer the rapid growth
             | 
             | How are these industries going to absorb new headcount
             | without the revenue to support it?
        
         | danielbln wrote:
         | You will always be able to produce artisanal hand-set code,
         | same as how artisanal woodworking exists alongside industrial
         | manufacturing. There will be a lot less demand for it, and
         | compensation will align accordingly, but it won't go away.
        
         | dude250711 wrote:
         | Either the crap truly works and nobody is needed or it does not
         | work. Where is this half-arsed human-agent hybrid vision coming
         | from? The land of plateaued LLM gains?
        
       | theelix wrote:
       | Moonlight-Web? I guess it's
       | https://github.com/MrCreativ3001/moonlight-web-stream but there's
       | no information in the article
       | 
       | > Moonlight expects: Each client connects to start their own
       | private game session
       | 
       | Nope, it's a Wolf design choice, eg. Sunshine allows users to
       | concorrenly connect to the same instance/game
        
         | lewq wrote:
         | Wolf now supports multiple clients connecting to the same
         | session via the wolf-ui branch that landed recently. After lots
         | of stability work we are now running that mode in production
         | (and in the latest release)
         | https://github.com/helixml/helix/releases/tag/2.5.3
        
       | _pdp_ wrote:
       | IMHO, the goals is not to have to watch what agents do and let
       | them do the work.
       | 
       | I would personally invest into making agents more autonomous (yes
       | hard problem today) then building a desktop video session
       | protocol to watch them do the work.
        
         | lewq wrote:
         | Yeah that's what we did :)
         | https://youtu.be/vVmnpcnLDGM?si=b6LxW6lmM7843LY0
        
         | majormajor wrote:
         | >I would personally invest into making agents more autonomous
         | (yes hard problem today) then building a desktop video session
         | protocol to watch them do the work.
         | 
         | Seems difficult to research better autonomy without extensive
         | monitoring. You need specific data on before/after effects of
         | changes, for instance.
        
       | lewq wrote:
       | Author of helix code here. Here's a demo of the full system
       | working. https://youtu.be/vVmnpcnLDGM?si=b6LxW6lmM7843LY0
       | 
       | We're opening the private beta where we provide a hosted
       | environment for testing, or you can install the latest Helix
       | release and run the installer with --code to try it on your own
       | GPUs
        
         | lewq wrote:
         | Join our discord for the beta
         | https://discord.com/invite/VJftd844GE
        
         | lewq wrote:
         | https://github.com/helixml/helix/releases/tag/2.5.3
        
       | reactordev wrote:
       | Whilst impressive to "bend a protocol to your will", why did you
       | not just take Moonlight and build on top of it, making your own?
       | 
       | No shoehorns needed. Just take what you like and build what you
       | need.
        
         | lewq wrote:
         | It's nice for unmodified moonlight clients to be able to
         | connect - they have tons of them, you can even run it on a
         | Nintendo DS
        
           | reactordev wrote:
           | But is the ability to run it on the DS a feature? I highly
           | doubt it.
           | 
           | I'm not trashing anything, I'm just saying that if they
           | focused on what their market is, it would be clear no one is
           | going to be coding/working on a Nintendo DS.
        
             | lewq wrote:
             | I suppose but we got it working, and the primary interface
             | is webrtc in the browser, and going via moonlight
             | internally is just an implementation detail that got us
             | here quickly. We are open to refactoring in the future of
             | course :)
        
       | CuriouslyC wrote:
       | This is beautiful madness.
        
       | momocowcow wrote:
       | What's the most intricate system that's been written with this?
        
         | lewq wrote:
         | Itself, more recently
        
       | luizfwolf wrote:
       | Hi, quite interesting project but have a hard time to understand
       | why would stream a desktop.
       | 
       | From my (ignorant) understanding, the important part is the
       | context of the LLM in the task. Some conversations you need
       | visuals, some you don't. What's the advantage of giving a full
       | desktop streaming instead of using integrations?
        
         | lewq wrote:
         | Mainly so you can give the agent access to the desktop as well.
         | Then it can debug your web app in Chrome Dev tools but also you
         | can pair with it with streaming that is so good it feels local
        
         | lewq wrote:
         | There's also value in being able to run multiple agents in
         | parallel with their own isolated filesystems and runtimes. One
         | agent won't tread on the toes of another whatever they do. You
         | can let it loose and it doesn't matter if it breaks something,
         | you can just spin up another one
        
       | vladgur wrote:
       | I'm curious how far are we from giving coding agents access to
       | these desktop agents so that when we are using say Claude Code to
       | build a native desktop app, the coding agents can actually see
       | and act on the desktop UI that it is building
        
         | drphilwinder wrote:
         | This is a great point. Not that far. We also snapshot the
         | desktop for "slow" non-streaming updates to the UI. We could
         | push these into Claude itself to act on or describe or
         | whatever.
        
         | lewq wrote:
         | That's the next move :-D
        
         | jsight wrote:
         | For web apps, I'd guess that many of us already do that via
         | Playwright or other MCPs. I'd bet there are people doing
         | something similar with desktop apps too.
        
       | kaspermarstal wrote:
       | > The Wolf maintainer has done heroic work ...
       | 
       | I commend the fact they acknowledge the maintainer's work, but
       | seeing the singular 'maintainer', I can't help but notice the
       | weight on that one person's shoulders.
        
         | lewq wrote:
         | I should have said creator. He seems to have a healthy
         | community backing him, but we should ask him!
        
       | vladgur wrote:
       | Another question regarding Helix - its being built as a platform
       | for private air-gap-ready ai agents that can work against private
       | LLM models.
       | 
       | Are there appliances or easy to deploy hardware that allow one to
       | run these private models on-premise vs cloud
        
         | lewq wrote:
         | Hey! Yeah we are working with partners on fully integrated
         | hardware+software stack for this. We particularly like the RTX
         | 6000 Pro Blackwell chips for this
        
       | LunicLynx wrote:
       | I find this very curious. I don't think agents care about UI,
       | humans do. So in the end the UI is not required. As soon as the
       | ai can get into the physical world. The whole IT world is done
       | for. All of this will be automated away. The IT and CS only ever
       | started to make us more productive more connected to improve our
       | physical well being. When we don't need to touch computers
       | anymore there is no need for ...
        
         | lewq wrote:
         | Vision language models have been trained on how to operate
         | human UIs though, so at least for a while, computer use will be
         | an interesting area to explore. I think debugging web apps and
         | building UIs is a particularly fruitful area for this
        
       | mxkopy wrote:
       | I've also independently concluded Moonlight was the best way to
       | go after trying my hand at a very similar task. I didn't want to
       | dig through moonlight's source, but I'm sure if you're dedicated
       | enough it would pay dividends later on, it basically does
       | everything you'd need for realtime control in the setting of
       | simulating human input.
        
       | jarym wrote:
       | Trying to do something similar but using kasm[0] as the backend.
       | 
       | [0] https://kasm.com
        
         | lewq wrote:
         | Fascinating, wanna compare notes on a call some time?
        
       | eisbaw wrote:
       | xpra has video streaming and allows for sharing
        
         | lewq wrote:
         | Interesting, thanks!
        
       ___________________________________________________________________
       (page generated 2025-11-15 23:01 UTC)