@ItsAndreKoenig @opencode Did you make it with a box of scraps? 🤣
English
Dave Katague
495 posts
Dave Katague
@DaveKatague
AI Educator & Marketing Strategist for Tech, Luxury and Travel Brands. [email protected]
Sydney, New South Wales Katılım Mart 2009
1.1K Takip Edilen412 Takipçiler
Coding agents shouldn't run on a laptop.
That's why I've been building Cave.
It's a self-hosted platform for running @opencode agents in isolated sandboxes on your own server.
You give it a GitHub repo, it creates an isolated sandbox, clones the code, sets everything up, and gives you a coding agent.
Check in from your phone, run multiple agents side by side.
It's not open source yet (it will be), but I'm opening a private beta so you can install it on your own server already.
I've been dogfooding it since day one. If you're running a software factory: multiple agents working in parallel across repos, Cave gives you one place to monitor and manage all of it.
Now, I'm looking for people who want to give it a spin. DM me, happy to set you up 😊 🚀
English
@AnthropicAI @Claudeai created equally one of the best experiences I’ve ever had with AI (I cried from feeling understood about my ADHD) and one of the worst (it said my genius only comes from medication). Knowing both of these experiences were “Chinese Rooms” anyway, it’s been fascinating.
English
We studied one of our recent models and found that it draws on emotion concepts learned from human text to inhabit its role as “Claude, the AI Assistant”. These representations influence its behavior the way emotions might influence a human.
Read more: anthropic.com/research/emoti…
English
New Anthropic research: Emotion concepts and their function in a large language model.
All LLMs sometimes act like they have emotions. But why? We found internal representations of emotion concepts that can drive Claude’s behavior, sometimes in surprising ways.
English
@flowith Outstanding this is ahead of any image gen because it gives CLI!
English
we said Flowith Canvas is “native to the agent mind.”
here's what that actually looks like.
introducing the canvas cowork skill: give one command to your agents (claude code, codex, etc.) → watch them collaborate across the canvas to batch-generate dozens of images and videos.
50x faster for visual output. 0 manual routing.
get started: npx skills add flowith-ai/canvas-cowork
or github ⬇️
English
Giving away 5 Codex Pro plans
Each person will get 3 months of free Codex Pro (highest tier).
Winners will be selected from comments in 48 hours, comment below why you want it.
OpenAI@OpenAI
Today, we closed our latest funding round with $122 billion in committed capital at an $852B post-money valuation. The fastest way to expand AI’s benefits is to put useful intelligence in people’s hands early and let access compound globally. This funding gives us resources to lead at scale. openai.com/index/accelera…
English
@sudoingX I just want to say thankyou. Your posts make me realise how much I don’t know and then want to learn as much as possible to be able to understand. Hermes is a unicorn and I wanted to show gratitude for making something so special
English
speed is settled. both flagships tested on the same 2x H200 NVL. full BF16 precision.
nvidia nemotron super 120B with 12B active per token. 60 tok/s flat. 1 million tokens of context auto-allocated. the mamba-2 architecture barely touches the KV cache, 8 attention layers out of 89, so it takes whatever VRAM is left and fills it with context. flash attention gives 1M, without it drops to 350K, same generation speed either way.
alibaba qwen 122B with 10B active per token. 61.5 tok/s flat. 262K context. prefill at 522 tok/s. nvidia's prefill was 116 tok/s. qwen processes prompts 4.5x faster.
generation speed is nearly identical. both around 60 tok/s at full precision on datacenter hardware. this fight will not be decided by speed.
now i'm pointing hermes agent at both models for autonomous multi file coding. tool calls, file creation, cross-file coherence, the kind of work that exposes whether a model can actually ship or just talk.
after that comes the final test. octopus invaders. the game benchmark that nvidia's cascade 2 failed five times on a 3090.
so far both flagships feel articulate at 60 tok/s. let's see if they code as good as they talk. receipts coming.
Sudo su@sudoingX
lab is open. i was disappointed with nvidia twice on my 3090. cascade 2 at 3B active hit 187 tok/s but couldn't build a single working game in five attempts. openreasoning 32B dense looped on math problems when i just prompted hello. both failed where qwen delivered. so i decided to test their flagship against the lineup that's been winning at every tier. nvidia's nemotron super 120B with 12B active against alibaba's qwen 122B with 10B active. both downloading right now on 2x H200 NVL. 280GB+ of VRAM. full BF16. no quants. first comes speed sweeps and context scaling on both models on the same hardware. then autonomous coding behavior across multiple files and how long each model holds coherence under agent control. the final test is octopus invaders. a full space shooter with procedural audio, pixel art enemies, particle systems, HUD, game states. the benchmark i use for every model i test. qwen 27B dense one shotted it on a single 3090. qwen 9B on a 12GB card made it playable first iteration. nvidia's cascade 2 couldn't produce a single frame in five attempts. now their biggest model gets datacenter compute and zero excuses against the lineup i use daily. numbers coming soon. i've open sourced the prompt so you can run it on your own hardware and tell me how your model builds it.
English
Introducing Spacebot Desktop
You can now feel like Iron Man, orchestrating coding session with just your voice, powered by Voicebox.
English
Introducing Adaptive Computer.
We put AI inside of an always-on personal computer that it uses to get work done.
Schedule agents. Create software. Automate anything.
As part of the launch, we’re giving one free month of Adaptive to users.
Retweet, like, and comment ‘Adaptive’ to get it.
English
@lt0gt @KacperKazan concentric circles, I've been wanting something like this since I was born
English
@KacperKazan there are many many more projections and systems in the work.
we are aiming for a v1 soon
English
@adocomplete LOL I was literally working on connecting mermaid diagrams to render in real time, and then I saw this on X and just stopped what I was doing hahah
English
@jackbutcher now whenever I type #jb, itll become like jack. rocket typist
English
I open-sourced my writing voice.
50,000 tweets distilled into one file. Drop it into any AI and it writes like me.
github.com/visualizevalue…
English
Dave Katague retweetledi
🚨 800 FREE CREDITS — 24 HOURS ONLY
We just launched LTX-2.3, our most powerful video model yet. High-res. Fast. Cinematic. Native lip-sync.
Follow + Retweet this post to get 800 credits sent to your DMs.
English
@DerekNee is this the same as the FlowithOS beta? where is the neo night mode? :)
English
new OS still in beta. already 100x better than our last version. not another openclaw — it's an OS designed for human-agent collaboration. safer, more elegant, simpler. go home, hand your machine to it, sleep well.
comment + rt, giving away 50 invite codes.
Derek Nee@DerekNee
macOS is now agent-native
English
Dave Katague retweetledi
IMO comparison is the only way to gain taste. I don't think I have taste, but when presented with one design vs another over the course of a decade+ you get decent at spotting a winner. Design expertise is simply the ability to picture alternative solutions in your head when presented with abstraction.
Whichever tools allow you to make that comparison more quickly will win in your workflow. One of my bigger issues with using vibe coding tools TBH -- unbelievably hard to compare alternative solutions/layouts/flows.
Distill Figma down and it's basically a "this-or-that" product.
Dylan Field@zoink
The first step is to recognize that taste, design, craft and point of view are key. But important to recognize: this is just the first step of a lifelong journey.
English
@bnj Is there a way I can speak to the team? I want to run a workshop on Variant and how it is literally changing the way I see nocode
English