
Founder @PolySimulatorHQ | Building the Standard for Prediction Market Backtesting & Simulation | CS @TU_Muenchen
Munich, Germany
Joined June 2022
- Tweets1.3K
- Following2.3K
- Followers252
- Likes23.3K
instinct is really the first agent where i can see AGI working.
i used to procrastinate constantly - not because i'm lazy, but because i know how painful some things are. government websites, forms, logins, all of it.
now it's the first time i never procrastinate. i just say what i need in the moment i need it and instinct does it for me. in 90% of cases without even asking.
i set up my bitwarden vault directly in its chrome instance. github cli is set up there. it has my obsidian vault, all my nodes, all the access it needs.
and it almost always just works by itself.
posted with instinct 😄
4
1
132
one thing i do like about muse: you can take control of the browser yourself. this is completely missing from instinct and it's one of the biggest features - i can solve the captcha, look at things
15
i tried both muse and instinct. and i can say instinct is just on another level: really understands you from one word, does exactly what you want, asks exactly at the right time. same task, two agents: muse asks me to enter the same password again. instinct just uses the other one. we need common sense agents
45
Wladimir Kirjanovs retweeted
NEWS: Boston Dynamics has just released a new video of its upgraded next-generation humanoid robot called Atlas.
• 4 hour battery. Self-swappable for continuous operation
• 6 feet 2 inches tall
• Weight: 198 lbs
• 56 total degrees of freedom
• Now fully electric, ditching older hydraulic systems
• New lightweight mix of aluminum and titanium components
• 110 lbs weight capacity (66 lbs sustained)
• Can reach up to 7.5 ft
• Constantly evaluates its surroundings and adjusts its posture, balance, and grip in real time
• Hands that can reconfigure as needed. Tactile sensors feed data back into the system, helping apply the right amount of force
• Brain is powered by Nvidia chips
Finally Live Translation that works
Today we’re introducing state-of-the-art text translation quality built with Gemini in Search and the Translate app for even more natural and accurate text translations. We’re also improving our language learning tools in the Translate app for practice and skill building.
7
476
Wladimir Kirjanovs retweeted
Comet now syncs your information across multiple desktop devices.
When syncing is turned on, Comet generates a unique sync code specifically for your account. Choose to sync everything, or customize the information that is shared across your devices.
Wladimir Kirjanovs retweeted
Gemini CLI is incredibly generous. With a $20 Google AI Pro plan, you get 1,500 requests per day (!) in Gemini CLI, which is essentially unlimited.
This includes Gemini 3.0 Pro, which will be released next week.
Unlimited vibe coding for 20 bucks!
Bought me perplexoty max. Will try it out tomorrow especially the email assistant, as I am overflowing...
Wladimir Kirjanovs retweeted
What AI tools are actually changing how I work? Here’s what’s on my desk (and in my workflows) this week 👇
@comet Browser - I'm all in now with Comet. Creating shortcuts for my most common prompts is awesome and being able to use them in the sidebar of any website is a game changer once you have a few solid workflows built. Earlier today, I had Comet review a spreadsheet of new AI tools, read the OG data from each of the websites, and pull in the name of the tool into a separate column. It's a simple task but I let Comet go off and do it for like 50 tools while I went off and worked on something else. It was awesome.
@warpdotdev - I slept on this one for a while. I assumed it was another coding tool. But it's actually a "use natural language to control your computer's terminal" tool. Have you ever had a file on your computer that you couldn't delete and couldn't figure out why it wouldn't delete? There's a terminal command that'll help you delete it... Warp does stuff like that. It CAN also code and you can run things like Claude Code from within it but it's much more than that. I used it to create an automation where, if there's an app in my computer's dock that I don't use for 30-days, it automatically removes it from my dock to keep things clean. And that’s just scratching the surface.
@n8n_io - If you've been watching my YouTube channel lately, you know this is one my current favorite rabbit-holes. I'm feeling like there's less and less that I can't automate with n8n. You can even tie it into smart home devices and create automations that control your home's lighting and things like that. The first time I used it, the learning curve felt steep but, after a few YouTube videos, I felt pretty dang comfortable building out workflows with it.
@Replit - Another one that I slept on for a while. It's currently my favorite "vibe coding" platform. I asked it to build a clone of a certain site recently. Just a super simple prompt. It built out an entire roadmap, executed on the roadmap step-by-step, double-checked it's own work to make sure things were functional, fixed a few mistakes, allowed me to preview it directly in Replit, and then pushed the whole thing to GitHub for me. It took maybe 15-minutes but I just gave it the prompt, opened a different tab to work on something else and, 15-minutes later, the app was built. (Disclaimer: I recently invested in Replit)
@WisprFlow - I'm trying to get in the habit of talking out my prompts. It still feels a bit awkward but I'm finding myself doing it more and more. Especially with Comet. Comet does have a built-in voice feature but it doesn't work nearly as good a WisprFlow. With WisprFlow I can just sort of rant into my microphone and the app will clean it all up, remove things like ums, fix any misspeaks and things like that before pasting it into whatever app I'm talking to.
(Disclaimer #2 - Some of these companies have sponsored my channel in the past or are sponsoring the channel in the future. But these are genuinely the tools I'm playing with the most at the moment)
At my core, I'm an automation and workflow guy. I got into YouTube to teach automations and workflows for entrepreneurs. AI was sort of the natural evolution from that. I wish my automation and workflow tutorials performed better on YouTube because I'd probably make nothing but those if I could. lol. But I gotta pay the bills and reporting on AI news is still what gets the views. So I have to make like 25% of my videos about automation instead and the rest is what attracts the eyeballs. Is what it is. :)
What tools are you digging right now?
Google promises it for multiple years, perplexity does it apparently. I even though about creating such a startup, but knew the giants will win either way and it is quite difficult to implement to be honest
BREAKING 🚨: Perplexity is rolling out Virtual try on feature on Perplexity for iOS.
“Try on” will allow users to generate their full body AI avatars to be used for shopping and trying clothes virtually. It will likely be US only.
99
Makes sense. Cheapest labor. And very important
My latest for @latimes: I travelled to southern India to document the rise of the AI "arm farms" — where young engineers strap GoPros to their foreheads and fold laundry or pack boxes to teach humanoid robots how to do chores.
latimes.com/business/story/2…
1
78
Wladimir Kirjanovs retweeted
I ran the same workflow in Comet, Atlas & FlowithOS.
❌ Comet timed out.
❌ Atlas failed at export.
✅ FlowithOS made the art, downloaded it, uploaded it to YouTube 🔥
This new agentic OS is invite-only.
I’ve got 10 codes. 1st come 1st served!
Just RT + comment FLOW in 🧵↓


























