Jose Domínguez

AI Support experience is broken, we want to fix it. I hate it when AIs send me a huge paragraph or redirect me to a help center. I especially hate it when I can't even see the button I'm supposed to click. Show, not tell try blar.io

@oneill_c This will still be very dependent on open source models. The RL loop only makes sense after pre-training, or do you see companies training models from scratch? My worries are when labs improve the baseline changes, and keeping up with that is hard (unless they flatten out)

@cacoos @facesdotapp Next, @Blar_AI promo video will be a @facesdotapp slide

hyperframes generates videos from code. Faces slides are made of code. So now you can prompt a video inside a slide! available as a community slide in @facesdotapp 🔊

@aakashgupta The question for me is this done via weight adjustment or simply re writing some MD doc? One feels like real learning, the other like re organizing your notes

Anthropic just shipped sleep into agents. When you sleep, your hippocampus replays the day's neural sequences to the cortex during 150-220 Hz bursts called sharp-wave ripples. The replay runs about 20x faster than the original experience. A 10-second sequence gets compressed to roughly 500 milliseconds. Wilson and McNaughton showed this in rats in 1994. You ran this algorithm last night on whatever you did yesterday, whether you wanted to or not. The replay does two things at once. It extracts statistical patterns: what mattered, what generalizes, which sequences predicted reward. And it reorganizes the memory trace from hippocampus-dependent storage into neocortex, which is why old memories survive hippocampal damage but recent ones don't. Disrupt sharp-wave ripples in a rat with optogenetics and the rat fails the next day's task. The replay is causal, not correlational. Most "agent memory" today is a search engine. Past sessions get embedded, you retrieve relevant chunks at the next call. That works for facts. It does not extract patterns and it does not reorganize the trace. Which is why agents plateau. The memory volume keeps growing while real capability flatlines. Dreaming reviews past sessions, extracts patterns, curates memories. That is the brain's actual three-step algorithm. They called it dreaming because dreaming is what the algorithm does, in roughly the same order, for roughly the same reason. Agents that dream between sessions will compound. The ones still running on raw context window will hit the same ceiling humans hit when they pull all-nighters.

@claudeai Is this memory curation done via MDs and RAG based? It would be interesting to have a LoRa adaptation dream state (much closer to human dream)

Dreaming reviews your agent's past sessions, extracts patterns, and curates memories so your agents learn over time. Request access: claude.com/form/claude-ma…

Live from Code with Claude: we're launching dreaming in Claude Managed Agents as a research preview. Outcomes, multiagent orchestration, and webhooks are now in public beta.

Today we are launching @​kapso/workflows: Build WhatsApp agents and automations, locally.

Day 10 of adding a new slide to my deck to show you what's possible with software presentations. Today, an interactive dashboard printed onto a piece of cloth, with a 3D fan blowing it around in real time. Fun. Link below!

@ignaciosoffia Niceee

@chetaslua @sama @facesdotapp @ignaciosoffia can we see wind simulation in an upcoming slide 🙏

HOLLLLY SHIIIIIT 😳 GPT 5.5 xhigh in codex made this paper physics website with wind effect one shot check the physics , ui and interaction @sama you guys cooked and this is my first time i can suggest chatgpt is go to ai solution for everything

meta-Faces

Day 4 of adding a new slide to my deck to show you what's possible with software presentations. This one is a web browser. Inside a slide!! You can embed a website and interact with it while presenting. Link below.

@ignaciosoffia Nice Faces inside Faces

@eliasmufa Yeah by all standards, if you showed this in the 60s - 70s, it's definitely AGI. The thing they are pursuing right now is smarter than human AGI, and for that, I believe the model still needs to learn as humans do, not a 1 time frozen weights.

@LePeppManu i think it is here already! we just got used to it

if this is not AGI i don’t know what is

@ignaciosoffia It would be cool to iframe a page in a pitch deck as "Demo" to send to investors. Faces inside Faces or this presentation inside itself recursive loop

Day 3 of adding a new slide to my deck to show you what's possible with software presentations. Created this very simple interactive testimonials carousel for a honey brand. Link below!

@toddsaunders What's your take on tools that help you guide this adoption? Many times B2B SaaS are bloated with features and hard to learn/use. We are creating an agent that helps users by guiding them

So many founders are doing this now. My inbox is flooded with stories like this.

@facesdotapp Amazing, been using Faces for a while, and I'm mind-blown by the product. Congrats guys!!

Introducing the new Faces. Interactive presentations that help entrepreneurs break through the static. If a website can do it, so now can your decks. Each slide is a software artifact, built for storytelling. Where others simply make deck-building faster or prettier, Faces are new interfaces for your idea, in its full glory.

@rauchg @adamwathan GOAT

Vercel will be officially sponsoring tailwindcss.com. That's a given. We as a community and industry owe @adamwathan and team a lot. Tailwind is foundational web infrastructure at this point (it fixed CSS 😉). I've also reached out to Adam to explore how we can make this a longer-term commitment.

@yoheinakajima @JayaGup10 Awesome, we did a similar project with Blarify a while ago as well. Trying to capture code in different nuances and abstractions. medium.com/blar-io/concep…

@JayaGup10 great article! i had this idea of a log graph that ties to a knowledge graph that feels similar to your tracing/context graph relationship.