Business Development Manager @ Infodesign Varaždin - AI practitioner and advocate - OpenEdge professional - Occasional tech speaker - F1 enthusiast
Republic of Croatia
Joined August 2009
- Tweets1.9K
- Following285
- Followers145
- Likes1.2K
Vjekoslav Jug retweeted
Incredible. I already have @Cursor's answer as to why the limits are disappearing.
Users who have Cursor Ultra through the SuperGrok Heavy subscription have had their "Other Models" usage pool reduced from $400 to $100. A 75% cut overnight. Without warning. Or at least I hadn't noticed. In fact, I could swear I read somewhere that the pool was now $500 and I was happy about it, until today. Has anyone else heard about this? The best subscription in my stack and my favorite harness are now useless. What a disappointment. @spacexai , @grok
Can you guess which models where used here?
Used @cursor_ai and @devindesktop as harness
Models that are used in alphabetical order are:
- Gemini 3.8 Flash (Cursor)
- GLM 5.3 (Devin)
- Muse 1.3 (Cursor)
- SWE 2.0 (Devin)
In post the order is different ;) Which one did best job for you? @borisplavljanic @lotar_vk
Things be like:
(with Fable)
Me: "Hi!"
Fable: (you used 13% of your weekly quota)
(with Grok)
Me: (working whole day on multiple programs, reports, writing documentation,...)
Grok: (you used 2% of your weekly qouta)
Okay, I am exaggerating a bit for dramatic effect but it is not far from that
Vjekoslav Jug retweeted
For the past two months I've been building a new AI benchmark from scratch, based on real tasks from my own workflows across fields that current benchmarks barely touch.
The goal is a different read on model capability: strict evaluation, detailed explanations, and clear results on which model is actually best for which kind of work.
It stays 100% neutral and fair regardless of sponsorship.
I've already put over $12000 of my own money into testing the major models.
If any company wants to sponsor the project or provide credits, that means more models tested, more trials run, and a more rigorous final benchmark.
Vjekoslav Jug retweeted
multica.ai 理念十分先进，很适合中小团队管理任务——可以分配给人或者多台机器上的 agent，还支持各种常用的harness。
对个人用户也极其友好：哪怕只当成看板做异步任务管理，配合我那套"种菜收菜"模式，效率同样拉满。
搞不懂为什么一直没爆火，真心推荐，开源免费，我不认识创始人。
xitter.cf/LinearUncle/status/208…
Vjekoslav Jug retweeted
More screenshots in this post.
multica.ai/docs/tutorial
Vjekoslav Jug retweeted
Introducing Multica
A harness for managing all your harnesses.
github.com/multica-ai/multic…
Okay, so many people seems to not like @AnthropicAI Opus (5) lately but it is actually not so bad as you guys saying, at least not for design.
Here are the results from Fable, Opus and Sonnet for making 3D model and Opus is far the best.
Made similar test with other model (except Kimi 3 - still on waitlist) and none is even close to Opus
Follow up on @OpenAI GPT results, here are the screenshots from models Sol, Terra and Luna made, all on max effort
I can post @OpenAI later, asked GPT to prepare it but can’t open on mobile. It did finish much much faster.
When comparing results I asked Claude why stronger model have worse result and it said it did focus on other points and not details so I gave Fable another chance (it said it will use more dense layout and put more elements). Still not comparable to Opus
Zadnjih dva dana aktualna je priča o psiću Bomboni koji je nestao na Hvaru. Suosjećam sa vlasnicima jer znam kako je to kad ti nestane ljubimac, no u toj priči mi je izrazito problematičan način na koji je nestao - “vlasnica apartmana je ušla u iznajmljeni prostor kad su gosti bili vani…”
Možda formalno ima pravo, ali između ostalog baš zbog takvih stvari izbjegavam domaće iznajmljivače i radije bih platio više samo da nemam posla s takvim ljudima. @Croatia_hr @HrvatskaNaDlanu @visithvar
Vjekoslav Jug retweeted
If you're using Linear/Jira and are heavily into AI, try Multica. It will blow your mind.
It's free and open source.
github.com/multica-ai/multic…
I've been seeing a lot of buzz (haha) about buzz.xyz and I do think this is where the human > agent coordination layer is going but I wanted to give a shoutout to the tool I've been using for the past while to coordinate a fleet of Hermes agents + Claude + Codex + Kimi
Multica.ai is a free, open source tool that I can best describe as agent native Linear and it's primarily where I'm coordinating with agents currently
I liked the Hermes Kanban but I didnt like having to switch between 2 separate interfaces to chat with my agent and then go look at the Kanban (@Teknium plz add the kanban to the desktop app)
Multica works exactly like Linear, you can
- Create projects
- Assign issues
- Build Timelines
- Setup schedules, etc.
But the cool part is it has native adapters for Hermes, Openclaw, Claude, Codex, Pi, OpenCode, etc. so you can actually assign those tasks to agents and watch them work on it
Or you can talk directly to them via the chat window tab
It also works really well with remote agents running on a VPS or another computer. Install the CLI, it connects up and boom, you're assigning tasks to the Claude, Codex, or Hermes install you have in the cloud
I think that as time goes on it's going to become clear that the interface through which we interact with agents is as important as the agent's capabilities itself if we want mass adoption and people using it for real work
That's why iMessage agents like Poke are so popular, the average person shits themselves at the sight of a terminal window but everyone knows how to use iMessage
I think Buzz is a really cool project, it's in the early stages and it will definitely get much better over time. It's fully aiming to replace Slack and I think it definitely can in the future. Once it adds some features like a decent Hermes adapter and some basic project management I could 100% see myself fully switching over
But for now I really love Multica, especially as someone who was using Linear for a long time
I'm not sponsored, Multica is free and open source, and I highly suggest you check it out if you're looking for a new way to interact with your agents
Link in the comments below
Can you recommend a good alternative to @superdoteng?
super.engineering/
One that works on windows 🫣
Quick facts - Grok 4.5 is roughly 5x cheaper than GPT and 10x than Fable 5. And probably significantly faster than both
After using @Jira for years at one point I switched to @linear and I love this app but lately I replaced it with @MulticaAI as I work more with agents than with people.
I really love how @MulticaAI does agent orchestration and helps me be more productive.
Any other users here?