Lars Buur

@hicasamadim 72

bunu çözersen, IQ seviyen ortalamanın üstündedir. çözebilir misin?

@ryancarson And you can see the replay from Modal here watch.getcontrast.io/watch/modal-ru…

Word on the street is that everyone is going to be switching back to Opus when the new model drops. This is exactly why I use an independent agent lab like Devin for my main software factory. They're going to deal with that headache for me. There's no way you can move fast and reliably if you're constantly switching between Claude Code, Codex and insert-other-shiny-object-here.

@ryancarson My bet is still on github.com/ColeMurray/bac…

@ryancarson i do not understand why anyone would announce anything April 1.

This is awesome

@ryancarson @corazoncap @saramfoster @efosta @HamelHusain @shl Congratulations 🍾

Just closed our $2m seed round. We're building an agent for a very specific consumer pain point. Once we scale to more states, I'll share more. I'd like to thank our investors for putting their chips on us! Lead: @corazoncap Angels: @saramfoster @efosta @HamelHusain @shl @HenryLSchuck @thrashr888 @jheitzeb @SteveMorin @kkliman @i_am_brennan @zachtdavidson @LexSokolin @hillarycbush @usiegj00 @WindAddict Matthew Collins, Greg Smith and Jed Rhoads

@pbteja1998 I wish you would focus on building features for your missionsquadhq. I am happy for your success, but the fact that you need to contact support to change your claude api or login is ludicrous.

Feb was one of the best months for SiteGPT.ai So many customers have subscribed to yearly plans… We made over $30k in revenue even though our MRR is only $20k…

@tveskov Det er fjollet at pege på en helt ny feature fra claude og sammenligne, som peter også skriver. Jeg personligt har det bedre med at mine personlige data ligger på et device jeg ejer. Jeg er på mac økosystemet med iphone og macbook. Assistenterne kender automatisk min fokus m.m

Har ikke helt fattet hvorfor 'man' skal ha en dedikeret Mac Mini til Openclaw, mange af de halvsløje usecases lyder som noget der kan laves med fx det her...

@SimonHoiberg You should look into using ruvector instead of pgvector on PostgreSQL. Either directly on Postgres (if you really need it), or directly just as a service.

I solved OpenClaw's memory issue. (At least, this is the best solution I tried so far). And yes. Big surprise. I solved it using RAG. All major models can now take millions of tokens as context. But the issue remains... The undisputed number #1 reason your agent is stupid and forgets simple things is because you're bloating context. So here's how I solved it with OpenClaw. 1️⃣ I installed PostgreSQL + pgvetor I run my OpenClaw on a Hetzner server, so I installed the database + pgvector extension directly on here. 2️⃣ Create a search tool Ask your agent to create a search tool for itself. Every time you ask it to remember something, it should: - Label the memory - Create a vector from the label - Store the label, vector, and raw text in the database Every time it's asked something it doesn't know, the FIRST thing it should always do is to use the search tool. 3️⃣ Memory CRON/heartbeat The agent can write to its memory file on-the-go. Consider this short-term memory. On a scheduled CRON (or heartbeat), it should "flush" its own short-term memory and store it in the database. Now, on every new session, the agent has very little context. The most important one is the description of using the search tool to enhance itself based on the task it's given. ✨ Major upsides - MUCH better memory - MUCH smarter - MUCH less token-greedy 👎 Major downsides - More moving parts - Complex for non-devs - Ongoing maintenance Still. Benefits outweigh the cons here. If you REALLY want to use OpenClaw professionally, I recommend that you use this 3-tool combo as a base: - OpenClaw itself - PostgreSQL + pgvector - n8n (for API proxies/security)

@sonderby Jeg er nysgerrig efter dine erfaringer med at huske? Er det blot markdown filer? Og i så fald hvordan vil du dele relevante viden/erfaringer mellem agenter - måske på flere maskiner? Jeg prøver lige nu supermemory.ai men er ikke glad for at "dele" den viden.

@sonderby Du behøver ikke flere Mac minis. En nogenlunde moderne mini med en god chunk ram er et bæst. I stedet for egen kanban har jeg brugt dhh’s fizzy.do Jeg er enig. Meget spændende tider og jeg kan påstå at har været “med” længere 🤓

@thekitze Lol. I feels like they just bumped the version to mess with the competitors

lmao opus 4.6 might not be agi after all 🤡

@dhh For your Basecamp orchestration notes: I spun up OpenClaw and PI and use Fizzy.do for orchestration. I use Gmail, so each agent just got the invite link to Fizzy and then uses my email+agentname@gmail.com

We're all still fumbling in the dark with agents, but some patterns are emerging (and some disappear quickly again!), and it's our role as software makers to make use of it all along the way. Here are some quick notes from one of those internal sessions sorting it all at 37s.

@pbteja1998 @openclaw I am "a bit" older than you, but I fully agree. I found your guide by chance and found it valuable. For future readers, you could mention that everyone can replicate your setup using fizzy.do. They have a small cli that agents "get" and you share an invite URL.

Last time I was this excited about building things was 9 years ago… when I got introduced to coding — the idea of thinking something and through code bringing it alive. Now, after 9 years, with @openclaw I have that same excitement to build things again. The only difference this time is the time gap between thinking and it being live is now seconds, not even minutes anymore.

@dok2001 I found Huang Jensen´s recent predictions about taste being a vital skill really insightful, which is somewhat similar to your "be elegant".

In 2026, taste is the engineering differentiator. Building is easy now. Knowing what to build, and what not to, is the hard part. Listen to users. Be elegant.

@barrowjoseph How does this compare to Adobe Sensei

Link: detect.semanticdocs.org Source and Models: github.com/jbarrow/common…

Convert any PDF into a fillable form with 🪄AI magic. I built a free tool that: (a) takes seconds, (b) outperforms all the commercial solutions, and (c) is open source!

@v0 Profile: @larsbuur" target="_blank" rel="nofollow noopener">v0.app/@larsbuur V0 app: gff.wizflow.io I built this app because the global fintech fest agenda was impossible to navigate and use effectively

To celebrate 100k followers, we're launching another round of "Free0" The first 1,000 replies that share an app they've built in v0 + follow the instructions pictured will get $100 in v0 credits. Good luck!

@paoloardoino Please remove the 1 number requirement for keet usernames. It blocks me (and others) from creating a username.

Keet now has username search and pairing. All P2P. No phones, no emails, no central servers, no intermediaries. 100% direct.

@paoloardoino @LuctorFrits @keet_io If I set my username to, for example, "myusername1", can I then remove the "1" later?

@LuctorFrits @keet_io At the moment to avoid name squatting until we iterate on the product

🚀 Keet 3.25 is here. Usernames have landed! Now you can: ✅ Claim your Keet @ “username” ✅ Search for other peers (We’re still working on being able to DM users directly via search, but it's coming soon)

@keet_io The 1 number requirement is silly if current online nick is without

@leeerob @philw_ @vercel_changes @v0 No zip code at all😀

@philw_ @vercel_changes Built a quick one with @v0 😁 v0.dev/chat/GiyVkMGZi…

Vercel Functions now include the requester's postal code. vercel.com/changelog/requ…