sav

hermes @NousResearch > openclaw let me explain... actually i don't totally know how yet. it just feels better. that sounds crazy until you remember @meetgranola raised $125m for an AI notepad. the moat isn't always a technical wall. sometimes its just a vibe thats hard to replicate.

floormaxxxing always

VC: You sure of the product? Me: Look at him, that's my founding engineer... Look at his eyes, I'll give you a hint, his name is @savboj. He's won major hackathons back to back - @GoogleDeepMind, @cursor_ai, .... Yeah I'm sure of the product.

People talk, listen, watch, think, and collaborate at the same time, in real time. We've designed an AI that works with people the same way. We share our approach, early results, and a quick look at our model in action. thinkingmachines.ai/blog/interacti…

Announcing the first OpenAI Codex Community Meetup in London. This Thursday 14th May @ Kings Cross we've got some special guests & surprises in store for this launch event get ready for some hardcore goblinmaxxing link below

i see the primitives forming.... pairing this with github.com/open-gitagent next... possibly the best memory system i've tried.... and the EASIEST to manage. makes you think how much of this stuff was over-engineered

5/ clean up > nothing worse than a bloated memory system > job runs a few times a week > creates backlinks > graphs > keeps organizational standards

4/ morning newsletter > while im sleeping > DEX navigates the web > finds AI news / Git repos / anything cool happening in the space > presents it like a newsletter > crazy alpha in there every morning

i've been using obsidian + agent harnesses for a while now, before @karpathy made it cool my core use-case has been: always-on memory layer:

Today @usemonk is announcing our $25M Series A co-led by @footworkvc and @AcrewCapital with continued support from @btv_vc. $3T sits in U.S. accounts receivable. Our mission is to close the gap between earning revenue and receiving it. For most B2B companies, that gap is 45-90 days, thousands of hours of manual work per year, and millions in trapped working capital. It's a tax every growing business pays. Monk invented the Intelligent Collections category. Most AR tools automate the easy parts. We built for the edge cases, portal uploads, PO mismatches, and disputes. On average, customers see 40% lower DSO, 24% higher collections response rate and 25+ hours/month saved. Thank you to @ElevenLabs, @tryprofound, @siro_app, and our customers for the trust. Our platform typically pays for itself in month one. Book a demo to learn more.

"devops engineering" is now writing, update-machine-playbook.md for agents to follow.

What I've been working on: I'm building heytavi.com with cofounders @VictorSunden and Moh and founding engineer @savboj Tavi is our vision of what the future of building teams looks like, but at it's core the idea is very simple: it's an AI agent that searches for candidates for you and sends you the results in your company Slack. But there's a bit more to it than that. Tavi is a fully-fledged @openclaw instance that sets itself up in under 3 clicks (we've counted) and in under 60 seconds. It runs on its own dedicated compute instance, and uses only frontier AI models (currently GPT-5.4). We handle all the billing integrations so it's literally zero-config. There's a good reason why we built it like this, as opposed to a simple chatbot, and it's because deep research tasks, like searching for hires, works best when you can marshal your whole company context, and have long-horizon memory over what does and doesn't work. We also realised that for small teams, hiring is a collaborative process. When you make a new hire, that person immediately becomes head of talent as you mine their network. So we built Tavi with that in mind from the start. What does this mean in practice? It means: - Metered billing. No per-seat pricing. The whole team gets to use Tavi from day 1, and we don't bill you when you're not using it; - Searches while you sleep. Tavi can do long-running deep searches and post candidates to slack drip-by-drip. If you don't like the results, just tell Tavi and it will adjust its approach; - Beyond LinkedIn. Tavi can enrich LinkedIn profiles with data from Github and personal websites, but it can also search our custom-built knowledge bases of special talent pools like hackathon winners, published AI researchers, and more; - Collaborative recruiting. Tavi can draft outreach and ping the appropriate member of your team to send the message. It's a great way to cut through the usual recruiter spam. Most of all, I'm incredibly excited about what the future will bring. Tavi is built on top of the bleeding edge of agentic harnesses but refined and crafted with enormous care and attention to detail to make sure that it does one thing extremely well: building your team. Right now, Tavi is free to try - no credit cards needed, all the LLM costs covered by us. (P.S. Tavi has its own X account @TaviNeverSleeps, I'm sure it would appreciate a follow)