Ex transportation tech-startup founder. Running on coffee and curiosity. explorer of anything ai.
Joined December 2023
- Tweets3.3K
- Following973
- Followers562
- Likes7.4K
Pinned Tweet
Hi @X people!
I'm 36 — ex-founder in transpo / mobility.
Last 3 years I've gone all-in on AI — agentic workflows, coding agents, open models.
Monday = time to expand my tech circle 🌐
Looking to build with:
- Founders / Indie Hackers / Startup Builders
- Vibe Coders / Developers / AI Builders
- Designers / Creators / Growth Operators
- Freelancers / Students
- Anyone shipping on the internet
If you're creating, learning, and actually building — let's connect.
Drop a Hey + what you're building right now 👇
🤖 Made with AI
Forgot to drop this. data-dash (claude dashboard) Self-hosted, provider-agnostic AI-agent skill for traceable offline HTML dashboards.
Sibling skill of my 'frame-motion' (claude motion) skill for non-claude users who would want to have it.
Forgot to drop this. Created a skill for my grok @bot but can be used on any ai agents.
It's a free alternative for claude dashboard. Self-hosted, provider-agnostic AI-agent skill for traceable offline HTML dashboards.
The sibling skill for 'frame motion', my free alternative for claude motion.
github.com/venthewise/data-d…
Forgot to drop this. Created a skill for my grok @bot but can be used on any ai agents.
It's a free alternative for claude dashboard. Self-hosted, provider-agnostic AI-agent skill for traceable offline HTML dashboards.
The sibling skill for 'frame motion', my free alternative for claude motion.
github.com/venthewise/data-d…
Claude motion is freakin amazing! But what if you're not a claude user?
I didn't wanna leave @grok but I got jealous.
So I made 'frame-motion'. This agent skill does the same job for free, in any AI agent: Codex, Cursor, Gemini, Grok, Claude, or whatever you're cheating on them with.
I use it with my Grok @bot and it made the video below.
Get your ai agent to just scan my repo and install it themselves.
👇 github.com/venthewise/frame-…
Happy to have 200+ new friends in a week! Being a reply guy is fun!
If you're into:
- Tech
- AI
- Building something amazing
Say "Hi" 👋 and tell me what you're working on! Let's connect!
A rare clip of @elonmusk in tears during Tesla's early days. (2016)
We see the flying Tesla Roadster today, but how was Tesla when it was just starting?
When asked if he ever worried about himself imploding: "No one, should put this many hours into work. This is not good. People shouldn't work this hard. They should not do this. This is very painful."
THE AI APOCALYPSE MAY HAVE A KILL SWITCH.
China gets the world hooked on cheap AI, then reminds everyone who controls the supply.
We thought we were downloading models. We may have been downloading dependencies.
Imagine waking up to find intelligence has become a geopolitical bargaining chip. 💀
Is China shutting off its cheap AI after getting the world hooked? Not yet, writes @cathythorbecke, but Beijing is reminding the world that it can (via @opinion) bloomberg.com/opinion/articl…
Read through the docs and got some questions to @Kimi_Moonshot.
1. Since you've intentionally capped Kimi 3.1 Pro’s offensive cyber capabilities, how does that guardrail scale under maximum context load?
(Eg. Can 900 pages of legacy spaghetti code break its alignment filters?)
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
@Kimi_Moonshot
2. With Kimi 3.1 Pro running 104B active parameters out of 2.8T, I’m curious at what point in a 900K-token window the MoE routing stops prioritizing compute efficiency.
(What I really mean: when does it stop trying and start guessing?)
Ok, so moonshot just quietly dropped kimi k3.1 pro and it’s kind of annoying how good it is.
Same size as k3 but suddenly trading blows with gpt-6 astra on coding, and sitting right next to opus 5.5 on some of the harder agent stuff.
Open weights nov 4. Already live if you just want to mess with it.
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
1M context, multimodal, 104B active. builds full games in unreal/unity/godot, does actual parametric 3D in blender and CAD, and they claim finance hallucinations dropped ~65%.
Highspeed version is 2-3x faster if you’re impatient. Artificial analysis has it jumping hard and looking like the new open-weight king once the weights land.
The gap is getting weirdly small.
xitter.cf/Kimi_Moonshot/status/2…
A rare clip of @elonmusk in tears during Tesla's early days. (2016)
We see the flying Tesla Roadster today, but how was Tesla when it was just starting?
When asked if he ever worried about himself imploding: "No one, should put this many hours into work. This is not good. People shouldn't work this hard. They should not do this. This is very painful."
Thought I might share this here.
Imagine, the same provider, different results? 👀
Grok @bot vs @grok Build using Grok 4.7 High
Did the Gummy bears challenge by one of my favorite AI creators @MiaAI_lab. What grok bot made aligns with the prompt given, unlike the one made by Grok Build.
Seems like what @elonmusk said about Grok Bot using models outside @SpaceXSI shines in this test, since it's very obvious that output 1 isn't made by a Grok model. 👇
Ventus retweeted
Imagine, the same provider, different results? 👀
Grok @bot vs @grok Build using Grok 4.7 High
Did the Gummy bears challenge by one of my favorite AI creators @MiaAI_lab. What grok bot made aligns with the prompt given, unlike the one made by Grok Build.
Seems like what @elonmusk said about Grok Bot using models outside @SpaceXSI shines in this test, since it's very obvious that output 1 isn't made by a Grok model. 👇
We went from training AI agents to genetically upgrading actual agents with paws. 🐕
The AI arms race is getting weird. Even the dogs are getting an upgrade before AGI arrives.
JUST IN: Palmer Luckey's genetically modified military dog startup, Warhound, comes out of stealth, aiming to create enhanced canines for the U.S. military.
Readers added context they thought people might want to know
This is NOT Palmer's company at all, his company Anduril is only supplying the tech for the dogs, and this is a wholly owed Collosal entiity
x.com/palmerluckey/s…
Hot take: Some of the most successful startups didn't eliminate problems. They created new ones that people are willing to pay even more to solve.
They make something 10x better, create 10 new problems, then build an entire industry around fixing them.
That's not a bug in capitalism. It's a business model.
Agree or disagree?
Imagine, the same provider, different results? 👀
Grok @bot vs @grok Build using Grok 4.7 High
Did the Gummy bears challenge by one of my favorite AI creators @MiaAI_lab. What grok bot made aligns with the prompt given, unlike the one made by Grok Build.
Seems like what @elonmusk said about Grok Bot using models outside @SpaceXSI shines in this test, since it's very obvious that output 1 isn't made by a Grok model. 👇
