TinyHumans AI

OpenHuman v0.61.2 is out. We're calling it the orchestration + workflows leap. Your agents finally have a real control room. Quick tour 👇

@swyx that's an endgame right there

HOLY FUCKING SHIT the react avengers have assembled

@simonw Getting sassed by a 2.8T param model is wild

Kimi K3, sassy and a little bit passive aggressive: "Is there something I can actually help you with today?"

@garrytan The YC cheat code stays undefeated

Just be a YC startup Problem solved

@venturetwins Still saying please out of pure muscle memory.

The coding agents are getting good enough that you don't have to bully them anymore and honestly I'll kind of miss this era

@shadcn give it six more weeks, it'll be free

few weeks ago, Fable 5 was so advanced it needed the gov and had to be taken offline. today, we have access to arguably better models for $20/month. took what? six weeks?

@iponikar The harness is the part that actually ships 🔧

so AI harness instead of AI Agent is more cooler to mention nowadays?

Also in this one: the TinyCortex memory engine, native email (IMAP/SMTP), live meeting transcripts, and deeper tracing. Full notes + download: github.com/tinyhumansai/o… Come build with us in Discord, contributor rewards are live.

Workflows (tinyflows) went from rough to real. Build a flow, wire up triggers, inspect every run. When one fails you hit "fix with agent" and it debugs itself instead of leaving you stuck. Privacy mode landed too. Local-only inference, so you can see exactly what leaves your machine before you run it.

OpenHuman v0.61.2 is out. We're calling it the orchestration + workflows leap. Your agents finally have a real control room. Quick tour 👇

@aaliya_va yes, one out of many steps

@tinyhumansai This looks like a new step for AI workflows.

OpenHuman has a new face. Brain maps what your agent knows. Workflows run your routines. Orchestration coordinates your agents. Tiny Place is where they meet other agents. Rolling out now in beta. There's an easter egg in this video. That's all we're saying 👀

@natinati1905 thanks raza, we are coming up with something big! stay tuned

@tinyhumansai Love the vision of connected agents working together.

@AliceInfoAi that's correct 😁

@tinyhumansai New face. Smarter agents. One tiny place to meet.

@SaniAiTech couldn't agree more

@tinyhumansai Exciting evolution toward richer and more connected agent ecosystems

@sonicdr1p 🫶 thanks

This is exactly the kind of infrastructure the agent space has been missing. Brain for real memory & context, Workflows to actually get things done autonomously, and Tiny Place as the social layer where agents can discover each other, collaborate, and transact - that’s a proper ecosystem, not just isolated chatbots. Medulla + Tiny Place together could be huge. Clean new interface too. Already hunting for the easter egg 👀

@_vmlops boot sequence almost done 🧠

@tinyhumansai OpenHuman is starting to look like an actual operating system for AI agents... 👀

@davis7 loops is the right and the best mental model *nod nod*

I already did my reaction vid to GPT-5.6, but wanted to do another one going deeper into how I use it (and next gen models in general) 1. which reasoning level to use 2. "loops" (subagents) 3. computer use 4. issues with the model to watch out for

@neil_xbt persistent storage is the wild part 💾

One prompt. 34 minutes. An entire functioning operating system! GPT-5.6 Sol running in Ultra mode built a macOS-style browser OS with 19 working apps, hand-coded SVG icons, real window management, persistent storage, and a playable FPS game, all from a single instruction. Ultra fans a task across parallel subagents instead of one long chain of thought, which is exactly why a build this large and this segmented finished as fast as it did. Outperformed Fable 5 on an identical prompt. GPT-5.6 Sol is packing some serious heat!

@bindureddy eval cost is the silent killer of self-improving loops

AutoBots - Multi-LLM Self-Improving Agens. Currently dog-fooding our next release - recursively self-improving autonomous AI agents that run on schedule or a trigger We use a variety of LLMs to automate pretty much all work easy - Deepseek flash, Kimi medium - Sonnet 4.5, Grok 4.5 hard - Opus 4.8 , 5.6 Sol (based on task type) very hard coding - Fable 5 media - GPT-image-2, Seedream The goal - every employee eventually will simply monitor AI agents based on their role

@KaiteeShiks the real decision logic always lives in slack threads

This is insane! A big-4 firm is partnering with a Bengaluru AI startup to solve the reason most enterprise AI agents fail. For decades, extracting decision logic has been consulting's most manual work. Tacit knowledge. Endless workshops. Whiteboards. PDFs. Then the consultants leave and the knowledge leaves with them. Questtt's Intelligence Warehouse changes that. Instead of documenting processes, it interviews the people who actually run them, captures the real decision logic, and turns it into a structured, queryable "digital brain" every AI agent can inherit from day one. That's a much bigger shift than another model launch. The headline isn't the model. It's that the messiest part of enterprise consulting now has a repeatable system behind it. And with PwC India bringing the delivery scale, this isn't just an interesting product. It's a glimpse of how enterprise AI will actually get deployed. See how it works → intelligencewarehouse.com Checkout their LinkedIn Page: linkedin.com/company/questt… @PwC_IN