#the Netherlands

1 messages · Page 1 of 1 (latest)

lost mica
#

I know theres a few of us around here. How about a hang-out in the coming weeks?

Gemini_Generated_Image_tuvk5ltuvk5ltuvk.png
#

If there's a lot of people I could organzie a meetup, but I've only so far seen a handful 🙂

#

cc @long quiver

long quiver
#

👋

#

sounds fun 🙂 where is everyone based? i'm in Rotterdam, Max you're in Amsterdam right?

lost mica
#

Amsterdam

worn thunder
#

I'm near Delft!

long quiver
lost mica
#

Damn I’m impressed with my nano-banana skills with the image above

long quiver
#

its pretty good, feels like i've been in that bar before. minus the lobsters

stray quartz
#

I'm in Eindhoven, but can move around

silver plank
#

Hi all, been exploring clawd all this week,
based in Amsterdam

latent glen
#

hey everyone, i'm in Amsterdam

eager pendant
#

I'm near Amsterdam too 👋

frank wharf
#

Howdy folks!
Utrecht here!

novel lynx
#

Amsterdam, sounds cool!

snow cradle
#

Groningen here, so I'm used to travelling 😉

broken palm
#

Maastricht here 🤓

snow cradle
#

Utrecht is looking like a good middle ground then?

stoic prism
#

Is a West-Flemish Belgian allowed?

long quiver
#

off-topic for meetups, but on-topic for NL: i yolocoded a new app (buienradar in menu bar) . yolocoded POS but will probably work with clawd/molt/???/bot easily if i make it a CLI. https://github.com/joshp123/buienbar (condolences to everyone in Groningen, the easiest place in the country to find rain today)

menu-bar.png
eager pendant
ashen blade
#

Gmgm ladies and gents; greetings from The Hague. Small time local LLM hobbyist. Reading up on Openclaw since a few days, trying to wrap my head around things, and considering getting started next week, assuming this is still around. So far, I gather that a dedicated workstation, whilst not required, appears to be best practice implied by real world experiences (the openclaw-fails thread in this discord has been very educative). Impressed by those firing this up on modest or vintage gear, but still leaning towards just getting a base model mac mini imagining this thing could soon be running 24/7 and to benefit from many users using that very same gear for a smoother overall experience. Next steps for me would be to figure out best practices for misc. accounts, have to see what is available, complies with terms and conditions, and is long term bot-accomodating, e.g. decentralized and permissionless, and whether supplementary risk mitigation measures (e.g. insurance coverage, antivirus) are warranted.

long quiver
ashen blade
# long quiver don't overthink it. just start with whatever you have on-hand, and don't give it...

Thanks for replying - don't get me wrong, I definitely see the potential here (was considering Manus before I learnt of this), just see some issues already and frankly worried that every Tom, Dick, and Harry firing this up on their old macbooks for proactive vibe coding might unintentionally end up hurting people. Yes, in the Netherlands we are very insurance-aware; maybe the next person from the Netherlands here will say that the answer to this is a "meldpunt" (government hotline or inbox for dedicated issue). Great that we have different perspectives, you never learn anything from hanging around people who just agree with your take on things, looking forward to getting started on this and an exchange over a beer or something one day!

raw zenith
#

Amersfoort, but no worries to move around

magic scroll
#

Utrecht here

brisk iron
#

eindhoven here

ashen blade
#

@mighty dragon @mortal lily - gmgm, saw you guys posted elsewhere, imagine we are all a bit new and eager to learn more, thought i'd draw your attention to possible upcoming real world meetup in the Netherlands. Apologies if this was out of line.

verbal ice
mortal lily
#

Almere here

ashen blade
mortal lily
brisk iris
#

Amsterdam!

vocal socket
#

Hardenberg!

snow cradle
ashen blade
# snow cradle I'm running my instance on a 2016 macbook pro that was laying neglected in a dra...

Thanks, that's probably sound advice - never heard anyone running LLMs complaining about overshooting on RAM, especially now in Ramageddon, which might last a while. I needed something dedicated so I ended up ordering one with 24GB that was on a deal for 1099 at Mediamarkt; figured it gives me a little more on-machine faculty at an attractive price point. 32GB or 64GB would have been nice, but felt a bit steep considering that:

  1. Mac Minis with M5 processor exhibiting impressive ttft and inference boost are expected to launch within a few months,
  2. I already have hardware running a llama.cpp server at home if it needs bigger models,
  3. Planning on developing an assistant that just does basic browsing and scraping tasks rather than coding or image/video generation,
  4. I am absurdly confident in my own ability to "prompt the shit" out of gear and receive above expected results on modest gear, and
  5. If I need access to SotA coding models, I am willing to get a dev account and pay for PAYG API access.
    This being said, I completely agree that if someone wants to try it out and has money to spend, and wants decent local performance, 32GB probably the right call.
snow cradle
ashen blade
# snow cradle Sounds good! I'm curious which models you're going to try locally.

Seems a 'receptionist configuration' is on the table, with a small model acting as a receptionist and deciding on how to route requests (i.e. load models on machine, or use APIs and in what order, so have to see how that is configured). For the receptionist some are using small Llama or Qwen models - I was recently positively surprised by LFM2 and LFM2.5 models from Liquid AI - they are aspiring towards cornering the forgotten market of potato PCs (sub 4GB RAM) by developing quite proficient small models (if you have decent phone, consider downloading their free apollo app for on-device private inference). If I had to pick something now for the receptionist it would be LFM2.5 1.2B instruction tuned but this really shooting from the hip - could be it is useless so would test against what others are using. For the main daily driver, guessing GLM 4.7 Flash, maybe Q6 right now REAP 23B Q4_K, but also shooting from the hip there.

sonic cloak
#

Hello 👋 Utrecht!

chilly plover
#

eindhoven here

ashen blade
#

Welcome @brittle musk @wooden dagger saw you guys joined, figure I'd tag you here - possible real world meetup soon.

wooden dagger
#

wonder if anyone here goes to AI Salon - I was there a month and a bit ago

brittle musk
#

hey, yes i would be interested as well!

#

thanks for tagging me 🙂

#

but hi everyone 🙂 i'm from Lelystad, can move around in all directions for a meetup 🙂

long quiver
#

now i can finally explain this to my gf's mum 😂

woven raft
#

Living in Nijmegen, working in Utrecht.

brisk iris
#

"Super Gevaarlijk", lekker Nederlands 😄

snow cradle
#

Is Alexander still relevant? I haven't heard anything about him for years now!

verbal ice
kindred flaxBOT
raw zenith
lost mica
#

I went to that event 👆 Was not really a builders event.

raw zenith
lost mica
#

A bit orthogonal to the proposal for a chill bar meetup, but I’m also working on organizing a meetup & hackathon via work (Picnic). Should also have some neat prizes. It would be in Amsterdam and probably on March 6th and hackathon ending on the 8th. I’ll share more details once officially planned but just giving the first fyi here.

mystic jewel
#

Amsterdam here. Has anybody here forked the code and if so what are you working on?

ashen blade
# mystic jewel Amsterdam here. Has anybody here forked the code and if so what are you working ...

Hi, welcome, really curious as to why you ask this - have you forked it yourself and are working on something cool? I barely know how to use a real fork, so forking anything is far beyond my skills, but I'm surprised anyone would consider forking openclaw right now given how fast things are moving, in particular with security. Guessing you are running something entirely offline? Or trying to strip away things? If you were asking about similar projects in general, I saw there was an 8 minute video posted on you tube yesterday called "OpenClaw vs PicoClaw vs NullClaw vs ZeroClaw vs NanoBot vs TinyClaw", dealing with some similar projects. I think he missed memubot btw.

mystic jewel
#

Hi @ashen blade, my experience with community projects is about zero and no I am not working on something cool. I would like to understand the code better so running a debug version locally would be helpful.
I want Openclaw to be more useful. I am not so much interested in making my own variant like Picoclaw, ZeroClaw etc. There are a number of essential improvements. In the current state, I won't give it access to my email, let alone my work email. I want to be able to use it more productively though. What I would want is that it technically can't perform risky tasks without my approval. Approval should be granted via an external process but light weight like with my finger print. Similar to when you pay in a shop with PayPal but with finger print confirmation.
If I can help the project in any way with that then I would be very happy.
In my experience it is amazing how small the group of actual contributors is, how approachable they are, even with a project that gets the world's attention like this one.

ashen blade
# mystic jewel Hi <@1164590034855415859>, my experience with community projects is about zero a...

Nice to hear from someone interested in business side use cases. Depending on risk appetite you can use openclaw in many ways, from full yolo mode, giving it access to everything, and sending anything to the cheapest ai inference provider on the planet, to using it offline, without ever connecting it to the internet in conjunction with local ai, which might be a way forward with confidential data. Lots of things are happening, and people are inventing new ways of using it, so trying to stay updated, tinkering a bit, and learning from the mistakes of others can be rewarding. There's a podcast called this week in startups where the last episodes have featured business side use cases involving openclaw, perhaps it is of interest for you.