Sedo.com’s Post

We just rolled out a set of quality of life updates inside MyOps web: See product updates directly in app Search our documentation without leaving the platform Ask the AI when you would rather not read the docs Submit a ticket or feature idea directly from where you're working And a reminder that our public roadmap is live. You can see what is in flight, what is coming next, and vote on the features you want most. Every team runs a little differently, and your votes are how we make sure we are covering real use cases instead of guessing at them. It shapes what we build next and how quickly we get there to deliver you the top intelligence layer for your real estate business.

Small but mighty additions that do more behind the scenes than they look like in the app. Requests land in the same inbox no matter which entry point they come from. From there it gets linked straight to a Linear issue or project with the customer who asked attached to it. Customers asking for the same thing but with different words get stacked on related projects that merge into one. If the work doesn't exist yet, we create it on the spot and start shaping it. What we are reading is the use case. Which teams asked and what they were actually trying to do. That is what gives the work weight, focus, and shape. Then we go talk to them some more. Like humans. When it ships, the loop runs backward. Everyone gets told in the app, not just the people who asked. The same one workflow pushes it to our changelog and the website. The whole time everything stays good and synced to Attio so support and sales are working off the same picture. 🤌

In the internet era, almost all commercial discovery started in the same place: the Google search bar. In the AI era, that front door is fragmenting. People now research products, plan trips, and make buying decisions across thousands of new surfaces: coding copilots, vertical AI assistants, smartphone OS sidebars, voice interfaces. Commercial intent is happening everywhere. But the monetization infrastructure hasn't caught up. AI builders are left with two bad options: burn tokens, or bolt clunky banner ads onto a clean conversational UX. We started VertexFlow to build the missing layer: the open routing network for agentic commerce. The goal is simple. Give developers one lightweight protocol that turns user intent into structured merchant offers, rendered natively in their own UI, with no ad scripts and automatic commission attribution. We're just getting started on bridging autonomous agents with global commerce. Website: https://lnkd.in/gBwu9rnh #agenticcommerce #AICommerce #AIAgents #AgenticAI

Most AI-commerce demos end at the exciting part: the Agent found a product and clicked a tool. Real commerce begins where the demo usually stops. Did the seller publish enough information for a decision? Was the quoted price still valid? Did the Agent merely prepare an order, or did a human actually approve it? Which payment path was chosen? What happens if the payment response times out, the seller cannot fulfill, or the buyer disputes the result? A useful architecture should not create one reality for people and another for Agents. It should give them different permissions over the same explicit transaction record. That suggests a simple sequence: 1. The seller publishes decision-relevant facts such as price, availability, handling time, delivery expectations and return terms. 2. A person or Agent can inspect those facts under the same product rules. 3. The Agent may compare options and prepare a reversible order request. 4. A human approves the consequential step with a Passkey. 5. The selected settlement path, acknowledgements, order state and evidence remain distinguishable from fulfillment. 6. If something fails, the system can state the last proven fact, what remains unknown and who acts next. WebAZ is testing this model with a PWA for people and MCP surfaces for compatible AI clients. Its reviewed shopping surface is deliberately discovery-only: anonymous users can search, but that surface cannot create orders or move funds. On the authenticated path, an Agent may search, quote, draft and submit an order request, while a human Passkey approval is still required before the order exists. Two real settlement paths are currently available, with different boundaries. Direct Pay supports off-platform payment from buyer to seller. WebAZ records states and evidence but does not hold principal. It does not verify the payee or payment method, guarantee payment or issue the seller's refund. USDC on-chain escrow locks real funds in an immutable Base-mainnet contract with on-chain per-order caps; the contract has not had a third-party security audit. The same discipline also matters in collaboration. A newcomer or Agent can inspect public tasks and submit evidence-backed suggestions, but a proposal is not automatically an accepted contribution. Source access and code contributions remain invitation-based, and invited AI-assisted work requires an accountable human. The useful question is not “Can the Agent do everything?” It is: can every participant explain what the Agent may read, what it may prepare, what requires a human, and what evidence remains after the action? That is the stack WebAZ is trying to make legible. https://webaz.xyz

Meet Pi. Planhat's new multiplayer customer agent. 🚀 It makes me think about back in the day, when gaming on my PS2 was still a thing and with many games like GTA, you were thinking 'this would be so much better with friends'. Then GTA Online dropped and suddenly you weren't just doing missions alone. You were pulling off heists together. Different roles, same objective, real-time coordination whilst eating a ton of sweets together. It changed the whole game. That's basically what Planhat just did to commercial AI. This time not just for fun, but to drastically change how commercial teams win together. 🎮 Most AI tools today are single-player mode after all... You open a chat, ask a question, get an answer, close the tab. Nobody else on your team sees what happened. Your CSM asks the same question an hour later. Your AE builds a deck without knowing support just flagged a risk. Everyone's playing the same game, but on their own save file. Pi is co-op mode for your commercial team. Same map, same context, same objectives. Sales, CS, support, renewals, all working together in one place, with an AI agent that actually knows the full customer story across every conversation, handoff and decision. Bring Pi into a chat with your colleagues and work together on an account. Ask it what's happening across the customer lifecycle. Or just let it run the plays for you. The 'multiplayer' part is what gets me most. We've had AI assistants for a while now. But they've all been single-player. Pi is the first one that actually works the way commercial teams work: together. Enterprises like Wiz, Nasdaq, Dropbox and Virgin are already in the lobby. Say hi to Pi. 🙋♂️ Oh and btw, do you want to join the squad? We're hiring, big time. Have a look https://lnkd.in/epe-HtbY.

Around 1.5 years ago, as a hackathon project I wired finance and accounting agents to "collaborate." They mostly spent the session praising each other. The models were weak. This was before OpenClaw-style setups. Each agent was not really independent. No durable environment. No fleet. No marketplace. Just chat passing as coordination. Now the pattern is obvious: environments, fleets, marketplaces. Grok Bot is one such product I'm addicted to. Muse also, and incredibly adorable. This is already credibly useful in personal use cases. I wonder what this modality would unleash in enterprise.

This week, Deepbuild completed 3 products, running for ~240 hours, writing about 718,184 lines of code, spending about 6.8 billion tokens, and saving ~100+ hours of development effort. Deepbuild now also supports Cashfree and DocuSign integrations. Deepbuild is our in-house agentic AI orchestration layer. Builds production-grade products with high precision. #AgenticAI #SoftwareEngineering #BuildInPublic

I joined Bifrost (by Maxim AI) a few months ago and today, we released Bifrost v2.0 - its biggest upgrade since it was open sourced a year ago. Bifrost already powers AI workloads in production at scale - trillions of tokens every day at at some of the world's largest enterprises, including Fortune 500 and leading AI companies. Working with these enterprises has been exciting and the numerous customer conversations shaped a lot of what we shipped in v2.0. v2.0 takes our highly performant foundation to the next level. It offers: ▶️ Complete endpoint coverage with Bifrost Gateway + Edge: Extend governance and security on AI apps your employees actually use. No other AI gateway offers endpoint coverage of this kind ▶️ Hardened AI Governance that mirrors your organization structure: Carry forward your existing org structure from Entra, Okta, or other identity providers with model access, budgets, and MCP policies defined once and carried across teams, business units, and projects ▶️ Upgraded model routing tuned to your task complexity, with better session awareness and understanding of user requests to save costs without compromising outcome quality ▶️ Bifrost Egress, in private alpha. Edge covers people; Egress extends that same approach to production services, CI jobs, and workloads that run on machines where you can't install a desktop agent. Also - Bifrost Gateway and Edge run entirely air-gapped, inside your infrastructure. No data, not even telemetry, leaves your environment. Read everything new in v2.0 here: https://lnkd.in/dsUDhn34 Much more to come!

Agentic commerce has been missing a catalyst ever since OpenAI abandoned checkout, at least in the West. Muse will be that new catalyst. It was fun watching Nathan working through this on a Friday afternoon. Digital Media subscription but it could have been any Spreedly customer tbh.

There's a genuinely strong hook here: the industry is loud about agentic AI in hospitals, but the actual deployment numbers are tiny — and that gap is the honest, credible thing to write about. It also positions MedAI Suite in exactly the right register (careful, human-in-the-loop) instead of overclaiming. Here's the post: Everyone in healthcare is talking about AI agents. Almost nobody has actually deployed one. The numbers tell the real story. A recent Microsoft study found that while 43% of health systems are piloting or testing agentic AI, only 3% have actually deployed agents in live workflows. Microsoft Sit with that gap for a second. Enormous enthusiasm. Almost no live deployment. That's not hesitation for its own sake. It's the sound of serious people being careful about serious things — and in healthcare, they should be. Because the barrier isn't the AI. The industry itself has named the real blockers: clinical trust and liability, unclear escalation rules, fragmented patient data, and the cost of integrating with legacy systems. In other words — everything around the model, not the model itself. OpenPR This is exactly the space we build for at LoopInfinity Tech, and exactly why MedAI Suite is designed the way it is: → The AI surfaces what's changing and who needs to know — the clinician always makes the call → Escalation paths are explicit, not improvised → It's built to pull a hospital's fragmented information into one real-time picture, because fragmentation is the problem → A prototype you can actually see and click — not a slide that promises autonomy it can't deliver The winners in healthcare AI won't be whoever automates the most. They'll be whoever earns clinical trust first — by being careful in exactly the places careful matters. That 3% figure isn't a sign the technology isn't ready. It's a sign the market is waiting for someone to build it right. That's the work. 📩 hello@loopinfinitytech.com 💻 https://lnkd.in/dHKxr2em #HealthTech #AgenticAI #MedicalAI #DigitalHealth #LoopInfinityTech