Sabitlenmiş Tweet
Find me on blue sky. More active there:
bsky.app/profile/onel.b…
English
Andrei Onel 🫎
926 posts
Andrei Onel 🫎
@AndreiOnel
building malmo, an OS for self hosting
Dublin Katılım Aralık 2011
321 Takip Edilen302 Takipçiler
I always wondered what the LM stands for.
Large Mongoose?
Lulu Mellon?
Language Molly?
English
@nickgomez @openknowledge This looks pretty good. What's the goal with it? Is it a side project that you guys are doing?
English
@OpenKnowledge is off to the races, almost at 3,000 stars (thank you!).
Since last week, we've added:
1. Antigravity, GitHub Copilot as built-in terminal agents
2. 1-click MCP install for LM Studio and Hermes
3. markdownlint and themes (!)
4. GitHub Enterprise Server support
+ tons of quality of life improvements.
Our goal is to the best markdown IDE and second brain for both humans and agents.
What do you want to see added?
English
We're starting to roll out early access for local dot ai.
We're 1% through the early access list.
We have a Local AI wall where anyone with early access can put up a post-it note. A few notes were already left on the wall. Once we release publicly, we're going to print it and frame it in the EXO office in SF.
Comment below and I will DM you an early access code.
English
This was a mistake.
New revenue created in the full test period came out almost 24% lower than baseline.
We reverted back to our 4 tiers exactly like we had them before.
Instead we're now offering new users to double their AI credits for the first 3 months - this seems to naturally push more new users towards the higher plans because the AI credits we offer to double are much higher.
Interesting experiment but didn't really work out.
Simon Høiberg@SimonHoiberg
We're experimenting with single-tier offer for FeedHive (instead of 4 different plans). Main reason is AI usage - we don't want to introduce usage-based pricing, it should stay simple. But to offer something that's truly useful and doesn't disappoint, the base price needs to be higher than the $19 or $29 we used to offer. Very interesting results so far! Fewer total conversions, but the higher price compensates. So MRR growth is unaffected (actually slighly positively affected), but we can serve fewer users better. This will probably be the new standard.
English
Some apps have their own login so you might want to make them public.
But for other apps, you don't want to fret about anyone guessing your url and accessing your private data.
English
You can now choose the access control for your installed apps in malmo.
You can make them publicly available, or only accessible to you.
#selfhosting #buildinpublic
English
Link to the discussion:
@mail.gmail.com/" target="_blank" rel="nofollow noopener">lore.kernel.org/linux-media/CA…
English
I think this is the advantage of projects or companies that have one or very few people guiding the vision.
The decision might not be the best but it's clear
English
Linus on AI:
> I realize that some people really dislike AI, but this is an area where I'm willing to absolutely put my foot down as the top-level maintainer. Linux is not one of those anti-AI projects
Very cool to see him talk so open about it but also to take a clear stance
English
@mitsuhiko This is my roman empire. I've been watching this since they were initially announced.
So disappointed that this is not the default way how we build web apps 😕
English
What's the current state of web components? Should we use it?
English
and less than one year later, we have it. and it's more magical than we ever could have imagined.
congrats to jony ive and the whole openai hardware team!
Sam Altman@sama
someday soon something smarter than the smartest person you know will be running on a device in your pocket, helping you with whatever you want. this is a very remarkable thing.
English
@pietergaricano I don't think I understand your point. If you want to talk to sales and I send you someone from HR, wouldn't you react the same way?
English
Unfortunately "Member of Technical Staff" did not resonate with European Parliamentarians
English
@liviusa @GergelyOrosz It's a social media website where you need to verify that you actually work at that company (by using the email address) in order to offer "scoops" about it
English
@GergelyOrosz Tf is a blind? Like a social media or what, first time hearing of such a thing. I still use hi5
English
I'll be posting a deepdive on this, having talked w sw engineers inside Anthropic's HQ about what they do.
The short of it: devs do a lot more+broader work than before; engineers do the design of systems (not AI); coordination v important. More to come in @Pragmatic_Eng
Santiago@svpino
Why is Anthropic hiring Software Engineers? Why aren't they using agents with loops to do all of this?
English
@thdxr What automations are you thinking about?
What if it runs in a completely safe environment where there's no risk of injection? Could they be enabled just then?
English
@bentlegen @sentry True. What are some good open source options?
English
we used feature flags very early on at Disqus and my entire time at @sentry (and again now at Modem)
they really should be dirt cheap though
Andrew Qu@andrewqu
I don’t care how small your startup is If you have real users, you should be using feature flags
English
@0interestrates A little bit, yeah. People want ideas from other people, you kinda loose big parts of that connection when AI is in the middle.
English
no escaping the sim rig lol
David Cramer@zeeg
this def aint podcast vibes.. gotta figure out if i use a diff room (ugh) or tweak how id use this one
English
@thinkymachines So awesome to see more open source models in the field 🎉
English
Today, we are introducing Inkling.
Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.
thinkingmachines.ai/news/introduci…
Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵
English
This is an interesting find.
Not sure how much to read into it though.
Is the test that needs updating or enforce those limits on the scores?
Tom Greenwald@tomgreenwald
@alexgshaw @pashmerepat Pretty widespread problem esp. from open model labs it seems. The widely reported 81% for GLM 5.2 was done with timeout=4h. It makes its way into a million graphics with no asterisk or disclaimer and all truth is lost
English