Vladimir

918 posts

Vladimir banner
Vladimir

Vladimir

@VladBrejcha

https://t.co/iexL35dYpi | @chattyfit | https://t.co/ydsNz8u3Mf Ask me about: AI Agents • Nomad lifestyle • Taiwan • Entrepreneurship

Taichung, Taiwan Katılım Mart 2025
114 Takip Edilen151 Takipçiler
Sabitlenmiş Tweet
Vladimir
Vladimir@VladBrejcha·
Okay there are too many comments about “vibe coding” and “ai coding” Most of them are out of context of AI powered companies 2026 reality: - if you want to be x50 faster and smarter you have to find a way to leverage AI - Vibecode IS NOT the way, as it creates more tech debt than value. - real ways: 1) build more internal tools for specific jobs; 2) orchestrate agents; 3) build long running pipelines; 4) evaluate and improve 50000times Only then you can expect agent to work better than human
English
3
0
6
314
Vladimir
Vladimir@VladBrejcha·
@hthieblot Not one person 5-10 different persons Obsessed but with no money and no contacts Capable of driving long run, those who admire difficulty and wants to beat the game no matter how hard it takes
English
0
0
0
18
Hubert Thieblot
Hubert Thieblot@hthieblot·
You just became a VC. You’ve got $1M to deploy. Who gets your money? Tag them. Or back yourself.
English
197
3
244
13.9K
Anthony Crognale
Anthony Crognale@acrognali·
@sindresorhus Which specific interactions are being slow for you now? We have a built in performance trace tool in the Help menu- would love to take a look if you send me the ID
English
3
0
7
1.2K
Sindre Sorhus
Sindre Sorhus@sindresorhus·
Oh how I wish the Codex app was native. It's so slow lately... Why can they not just make Codex rewrite it in Swift?
English
38
2
214
45.2K
Vladimir
Vladimir@VladBrejcha·
@bradleyleaveem @sindresorhus Chat is super laggy in codex app now It’s because of virtualisation not handled actually In swift its a basic primitive that is working efficiently out of the box
English
0
0
1
22
Bradley Gibson
Bradley Gibson@bradleyleaveem·
@sindresorhus The grass is not greener on the other side Apparently virtualisation is better handled in a web framework The app might feel better but the chat would feel jank
English
1
0
0
956
Vladimir
Vladimir@VladBrejcha·
@jaredsuniverse Life is one Can’t miss it And can’t miss being obsessed with something too
English
0
0
0
6
Jared Stivala
Jared Stivala@jaredsuniverse·
what kind of founder are you? 1. obsessed till something works 2. work life balance
English
88
5
94
7K
DeepakNess
DeepakNess@DeepakNesss·
@VladBrejcha the final outcome is good though, i have tested it on multiple real stuff and it costs me less as well
English
1
0
0
40
DeepakNess
DeepakNess@DeepakNesss·
42M tokens cost only $0.50 when using DeepSeek v4 Pro (Max). 🤯
DeepakNess tweet media
English
21
8
327
14.6K
Fraser
Fraser@iamfra5er·
Avoid building startups here - Cultural aversion to ambition - Insane regulation - Talent shortage - Out of control taxes The next generation will build their companies elsewhere!
Fraser tweet media
English
556
284
2.5K
168K
Vladimir
Vladimir@VladBrejcha·
@LuminousTheReal @Dimillian You don’t set it up manually You ask your ai agent to set it up Good start will be sending them openai articles about harness and asking “how could we leverage this to build our harness”
English
2
0
0
28
Thomas Ricouard
Thomas Ricouard@Dimillian·
The hardest thing in the Codex app is to know if a feature shipped or not
English
17
1
246
14.1K
Vladimir
Vladimir@VladBrejcha·
@jbhuang0604 The problem with deepseek is that these 10M tokens are 3M useful tokens and 7M tokens thinking in circles and making mistakes and rolling back In other words its fucking slow because its not smart enough Still good for isolated small tasks like evals
English
0
0
3
461
Jia-Bin Huang
Jia-Bin Huang@jbhuang0604·
Keep getting rate-limited by Claude, so I tried out DeepSeek V4 for the first time. After 10M+ tokens, holy crap the cost is ... 🤯
Jia-Bin Huang tweet media
English
203
201
5.1K
463.4K
Vladimir
Vladimir@VladBrejcha·
@anumness I’ve made the whole video course with lessons using only elevenlabs, heygen and lessons generates from templates Only screen recording was manual because I needed to show the actual coding
English
1
0
1
342
Anum 
Anum @anumness·
OMG this actually worked 🤯🤯🤯🤯 The database is complete now so I told Claude to use it to make a video of most viral moments with narration and the result is SHOCKINGLY AMAZING We are talking a full long form video (15+ minutes) with narration, motion graphics, clips from the show. A ready to upload video. I didn't expect it to work THIS WELL.
Anum @anumness

I'm in a rabbit hole again. While creating the faceless youtube content for my app, I discovered several youtube channels that are raking millions of views simply making faceless videos with clips + AI commentary from popular T.V. shows in that niche. And I've realised how much value is in such a channel. For example if your app is a dance app, the youtube channel could make videos like "The Best Strictly Come Dancing Performances of Each Year" But the trouble is collecting that information. How do you find that information and verify that its correct. And so Im building a database for the most popular show in my niche. Which is crazy but once I have a database which is mostly a database of transcripts for all episodes, it would be easy to ask claude to find me best/top 10/most shocking kinda content, find perfect clips to include. Imagine having all of that automated. It would be insane. Or maybe I'm insane for thinking its doable.

English
17
4
142
36.7K
Vladimir
Vladimir@VladBrejcha·
Gym day
Vladimir tweet mediaVladimir tweet mediaVladimir tweet mediaVladimir tweet media
English
0
0
1
41
Vladimir
Vladimir@VladBrejcha·
@nxhaaa19 Type 3 Builds, invests and talks daily 0 users 0 post impressions
English
0
0
0
9
neha
neha@nxhaaa19·
2 types of founders: Type A: Builds for 6 months, launches, 0 users Type B: Posts daily, talks to people, launches to a waiting list Which one are you? 👇
English
87
3
101
2.8K
Vladimir
Vladimir@VladBrejcha·
@BenittoJD @Italianclownz Its screenshot from deepseek api dashboard I use deepseek mostly in opencode with custom harness + in my apps
English
0
0
1
44
Carlo
Carlo@Italianclownz·
I tried Deepseek V4. And in less than 5 minutes I already burned .05 using opencode and vibe coding.
Carlo tweet media
English
64
6
753
79.5K
Vladimir
Vladimir@VladBrejcha·
Dm for a ticket
English
0
0
0
9
Vladimir
Vladimir@VladBrejcha·
Got 3 more tickets for sharing My events: 1. One founder may agents 2. AI founders and builders circle 25-25 May Taiwan Digital Fest
English
1
0
0
158
Vladimir
Vladimir@VladBrejcha·
@hthieblot That’s the lifestyle we choose basically
English
0
0
0
11
Hubert Thieblot
Hubert Thieblot@hthieblot·
If a founder says: “I’ll die before I quit.” “I have no plan B.” “I want to retire my parents.” “This is my life’s work.” Invest in them immediately.
English
148
50
1K
90K
Joe Stacks of Raverpay
Joe Stacks of Raverpay@joestackss·
Opencode x Deepseek v4 pro x High 800 million tokens = $10 The Chinese can have all my data
English
99
263
5.3K
160.5K
Vladimir
Vladimir@VladBrejcha·
Ok I’ll be honest about DeepSeek v4 - it is really as smart as gpt 5.2-5.3 - it is 80% as good for long running tasks (>1h) - It is practically free compared to others - BIGGEST problem: it very often IGNORES agents file. The tooling and harness therefore must be enforced.
Vladimir tweet mediaVladimir tweet media
English
0
0
3
1.2K

Keşfet

@hthieblot @acrognali @sindresorhus @bradleyleaveem @jaredsuniverse @DeepakNesss @iamfra5er @LuminousTheReal