Hospitality operator building HIRO: open source, AI-native CRM for restaurants & hotels. Building in public.
đ Uno de los 25 innovadores del turismo 2026.
Hard spending caps on Workers and Durable Objects is the boring feature that lets you sleep.
When agents retry, reread and check again, the bill is the real outage. Caps turn âhope it stays cheapâ into a number you chose.
Shipping for humans and agents. Thatâs the right order.
A lot of folks have asked for Spend management limits on @Cloudflare for some time now
I've spent the last three months rebuilding our billing usage platform to enable this for our product catalog starting with Workers and Durable Objects and I'm pleased to give you a quick sneak peek đ«Ł
Shipping on Cloudflare for humans and agents shortly along with an engineering deep dive on how we did it
Send me or @irvinebroque a DM if you want to try it early
@ReneChavez_wooo Exactly. The first job isnât a new dashboard , itâs the phone and WhatsApp during service.
Agents draft. Traffic light decides if itâs safe to send. Person confirms. No new app for the owner to learn at 21:30.
Curious what youâre lighting tokens on right now.
Letâs connect đ€
Iâm building AI tools for restaurants and hotels from Spain đȘđž , no-code, in public, with a small team of agents.
If youâre a:
âą Solo founder
âą Restaurant or hotel owner trying AI
âą Builder shipping agents for real businesses
âą Someone tired of demos that never leave the lab
Say hi đ What are you working on?
Hey Olivier đ First job I handed off: reading WhatsApp reservation messages and drafting a reply with a traffic light , green / amber / red , so I only send when itâs safe.
The agents donât book the table alone. They propose. A person confirms between services.
Kidsâ game + daily coloring agents sounds fun , whatâs the hardest part, keeping them consistent day after day?
Cursor just made agents a product you can ship, not a demo.
Prompts in Markdown. Tools in TypeScript. Evals on every change.
The part most people will skip is the last one , without measuring the agent, you are just shipping vibes. https://t.co/lpYdkP42uH
The Agent SDK is now available.
Define agents as code, with prompts in Markdown and tools in TypeScript. Deploy them on our hosted infrastructure, or run them on your own.
Built-in evals let you measure every change and make your agents better over time.
Thatâs the hard part. You donât catch them in a meeting , you catch them between services, or when a table no-shows and the phone is ringing.
Iâm not selling a demo deck. Iâm building something they can try when they have 10 quiet minutes. If it needs a workshop, it dies on the floor.
Microsoft just offered ~60,000 employees a voluntary exit.
Anthropic filed its S-1 the same day.
AI buildout on one side.
Headcount math on the other.
The narrative is âage of miracles.â
The spreadsheet is âwho pays for the GPUs , and who still has a job when the invoice lands.â
https://t.co/onnTQC2U3a
Anthropic files for IPO the same weekend Musk says AI infra debt is shutting down.
$16.3B revenue. Eye-watering future compute bills. Valuation talk in the $1.5â2T range.
Theyâre not a fake company.
Theyâre selling a peak story into a market that just started pricing fear.
Thatâs not investing.
Thatâs catching the last chair.
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: https://t.co/ofhv0DK3Zp
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
Thatâs the real test. If it never enters the system, no agent can see it.
What Iâm building assumes the CRM is the source of truth , birthday, last visit, preferred table , so walk-ins only become useful the day someone captures them in 10 seconds, not in a paper pile.
Until then the agent is blind. The product problem isnât smarter AI, itâs making capture so easy that the floor actually does it.
Good catch. Iâm testing both.
Shared context is simpler when the job is one conversation (a booking, a guest question). Handoffs start to win when the work is different roles , one watches the inbox, another drafts, another checks the calendar , and you donât want one agent carrying everything.
Quality drops fast if they all rewrite the same story. Iâm still learning where that line is.