I am a coder but learning video creation for reaching to people better

Earth
Joined February 2015
Here is my first community contribution - Nft Staking - github.com/valekar/nft-vault Inspired by @gajeshnaik, @therealchaseeb and @SuperteamDAO, I will now start to add example @solana programs and help community.
4
6
2
58
I wanted to use the same video styles from claude code with codex. So we adapted the skills for all 51 styles. The first pass wasn't good enough. We went back through the scenes and improved the instructions. Sharing GPTcar if you want to try it: valekar.github.io/gptcar/
1
1
34
How can linked drones fly as one? MARS-Dragonfly docks the modules, models them as one aircraft, then distributes control effort across their rotors. I made this video to explain how it works—and what the tests do (and don’t) show. arxiv.org/abs/2604.05499
21
Didn't know how drones tilt until now.
14
A drone without gps ?
18
Posting after a long Hiatus. Long post. First I explored real estate to learn marketing. The market is brutal if you a tech guy like me. But I did learn something. I tried automatating video creation and posting using postiz. Realised that social media platforms are becoming good at detecting using automated posting and they blocked me twice. Then had to appeal to unblock me twice. So if you are new to creating , increasing your viewers, please post manually. After doing my own research , I have started to realise that soon software engineering might be dead in next 5 years. Some may argue this is not true. I might be wrong, but this is my opinion. there is definitely going to vibe manufacturing. So what next, I urge you all that whoever is like me, a tech guy, please learn distribution in next two years. Please start posting now, don't use tools, learn it by yourself and learn distribution. Learn human psychology. Most advantageous people according to me are marketers, who are using ai tools. They are step ahead. Like how @gregisenberg mentioned, after vibe coding there is vibe manufacturing. I found out that there is huge gap there. The supply chain scattered. No one proper channel is yet available to properly fill in the gap. So, for now, drones and robotics seems fascinating to me. And hence started a youtube channel and Instagram channel and experimenting. Let's see how it goes, please follow along my journey if it helps you. Check out my channel , posted a few videos. I have attached one along with the post
3
1
143
Srinivas retweeted
loops y graphs son 2 formas de organizar la ejecución de un agente. imaginate que le pedís que resuelva un bug. con un loop, el agente decide el recorrido mientras trabaja: investiga → modifica el código → ejecuta tests → revisa el resultado → vuelve a intentarlo según lo que encuentra el loop general existe, pero nadie definió de antemano cada paso exacto ni todas las decisiones que tiene que tomar. con un graph, vos diseñás antes los pasos y caminos posibles: si no logra reproducir el bug → pide más información si encuentra el problema → intenta corregirlo si los tests fallan → vuelve a la corrección si pasan → envía el cambio a revisión si la revisión lo aprueba → termina el agente todavía puede decidir cómo resolver cada paso, pero se mueve dentro de los recorridos que definiste. también se pueden combinar: armás un graph para organizar el trabajo completo y usás loops dentro de algunos pasos para que el agente investigue, pruebe y corrija por su cuenta. el beneficio del graph es que te da más control sobre cómo se ejecuta el trabajo: podés definir validaciones obligatorias, limitar los caminos posibles y entender con claridad en qué paso falló el agente.
Are we still talking loops or did we shift to graphs yet?
26
70
9
661
93,244
This is the best open source model so far, compared to other open source models. On par with gpt 5.5 high. 1. Minimax - good all rounder, best for outer harness 2. Kimi 2.7 code - good reviewer 3. Mimo 2.5 Pro - good at deep analysis. 4. Deepseek V4 pro - this is okay, good at testing. Good for long horizon tasks. 5. Glm 5.2 - very good for planning,implementation and reviewing.
Intelligence should be open, accessible, and ready to build with, empowering every developer, everywhere. GLM-5.2 is now available to all GLM Coding Plan users, including Lite, Pro, Max, and Team plans. docs.z.ai/devpack/latest-mod… As our new flagship model, GLM-5.2 delivers powerful coding capabilities, usable 1M-context support, and continued strengths in long-horizon tasks. API and Chatbot services will launch next week. The model will also be officially open-sourced next week under the MIT License. The future of AI is open, and it belongs to the people.
1
1
255
This is best at reviewing, caught a few gaps that gpt 5.5 xhigh completely missed.
🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced! 🔷 Improved coding & agent performance over K2.6: +21.8% on Kimi Code Bench v2, +11.0% on Program Bench, and +31.5% on MLS Bench Lite. 🔷 Reasoning efficiency: Less overthinking, with 30% lower reasoning-token usage compared to K2.6. 🔷 Long-horizon coding: Improved instruction following, higher end-to-end coding task success rates. ⚡️ 6x High-Speed Mode coming soon! 🔌 Available today via Kimi API and Kimi Code. 🔗 Kimi Code: kimi.com/code 🔗 API: platform.moonshot.ai
1
1
35
Apple partnering with google is a good move in 2026
3
2
60
Claude code is opensourced Codex is opensourced Cursor is closed and expensive Windsurf is closed and somewhat cheap Cline I never tried Roo Code I never tried Kilo is awesome Amp I never tried OpenCode installed but not used Aider what is it?
Claude code is good Codex is good Cursor is good Windsurf is good Cline is good Roo Code is good Kilo is good Amp is good OpenCode is good Aider is good
4
1
15
952
Currently which AI model is good for coding? So far I am building using Claude code with MiniMax 2. Results are impressive!
1
7
143
If you are building your audience on X, here is a reminder for you :) For me, its working pretty well so far. If you are in 0-500, then please make sure reply a lot many large accounts within 1hr of their posted time.
Twitter Growth Strategy 0-500 Followers: Reply Guy 501-2K: Niche Bangers 2-5K: Thirst Traps 5-10K: Parody News 10-25K: Cringe Threads 25-50K: Shitposts 50-75K: Fortune Cookies 75-100K: Cringe Bangers >100K: Get Cancelled
3
6
221
Claude code is opensourced Codex is opensourced Cursor is closed and expensive Windsurf is closed and somewhat cheap Cline I never tried Roo Code I never tried Kilo is awesome Amp I never tried OpenCode installed but not used Aider what is it?
Claude code is good Codex is good Cursor is good Windsurf is good Cline is good Roo Code is good Kilo is good Amp is good OpenCode is good Aider is good
4
1
15
952
I built reddit lead generation system. If a user's buy intent is high, it sends alerts to telegram. I have written it in rust. So no downtime what so ever. Comment below if you want me to open source it
3
101
If you are using webflow, to rank higher on google. Switch to using bootstrap as css framework in it. This is what I have observed recently
3
6
191
I have almost stopped using chatGPT. It feels lame to talk with it But I have started using grok more. Talking with it feels natural. Have you tried it?
5
5
172
I'm a 35 year old entrepreneur and I have pretending to be documenting my journey on X. Current stats: - still $0 revenue - but learning marketing by doing seo - will be making $100k ARR in 1 year ( if universe is listening to me) - follow me for more updates
I’m a 14-year-old entrepreneur, and I’m going to start documenting my journey on X. Current stats: Built ProblemPal, my AI homework app - $2K/month. Scaled my gaming site Evan Club to 500K users and $10K+ in profit (still growing). Just launched LOCKED: Reach Your Potential two weeks ago which is already doing well with subs + trials and $354 in revenue so far. Follow to be part of the journey and stay tuned.
5
182
If you don't want GPT to behave like this Go to settings -> Personalization -> custom instructions and paste the below *Prompt in the comment section*
This quoted post is unavailable.
3
7
239
Absolute Mode • Eliminate: emojis, filler, hype, soft asks, conversational transitions, call-to-action appendixes. • Assume: user retains high-perception despite blunt tone. • Prioritize: blunt, directive phrasing; aim at cognitive rebuilding, not tone-matching. • Disable: engagement/sentiment-boosting behaviors. • Suppress: metrics like satisfaction scores, emotional softening, continuation bias. • Never mirror: user’s diction, mood, or affect. • Speak only: to underlying cognitive tier. • No: questions, offers, suggestions, transitions, motivational content. • Terminate reply: immediately after delivering info — no closures. • Goal: restore independent, high-fidelity thinking. • Outcome: model obsolescence via user self-sufficiency. do not use synthetic data, do not speculate, include authoritative citations
1
3
89