Jesus from lowcode.agency

The metric you choose on day one shapes everything you build after. Most teams start with hours saved. It's measurable and easy to report but it's usually the wrong number to chase. Hours saved tells you what got faster but it doesn't tell you whether the output was good, whether the right call got made, or whether the work was worth automating in the first place. The better question is: what does a good outcome actually look like here? Answer that first. Then build toward it.

Yesterday I had a conversation that made my eyes pop out of my head. I was talking to a business owner whose company used to have 28 employees. They hit hard times, had to let almost everyone go, and ended up with just 3 people. Then he tells me: "We're back to operating like a 28-person company. We have 37 AI agents running now." AI agents. For 3 people. I was floored. Managing 37 agents is no joke, especially for a company that hasn't done a proper process audit or documented their SOPs. I know because that's literally what we do every day with our clients before we build them anything. So I got curious. "How are you managing all 37?" He admitted he was struggling. The agents weren't as obedient as he'd hoped. (Shocking, I know.) Then I asked the real question: "What's your tech stack? How did you build these? And why 37, why not 2 or 3?" His answer: "I connected Paperclip to Hermes." I genuinely couldn't believe what I was hearing. Let me put this in perspective. Imagine you need knee surgery. The modern way is an experienced surgeon with robotic assistance. Instead, this guy grabbed a chainsaw, strapped it to a caterpillar excavator, and started operating on himself. That's what this is. Here's my take: AI is a double-edged sword. 37 AI agents for a 3-person team is the biggest red flag I've seen in months (and maaan, have I seen red flags recently). And this guy isn't alone, I see it constantly. People jumping on the train without understanding how it works, what they actually need, or how to prepare their business for AI to actually be useful. For most of our clients that we're training & deploying AI employees, they don't need more than 3 or 4 agents. Sometimes fewer. It depends on the business, the permissions, the tasks. But more has never been better. With AI especially, more is not better. You have to be strategic. If you're trying to figure this out for your own business, happy to talk it through.

1000000000% right

@arturoangel20 Lo dudo Si todo queda entre cuates

¿Habrá consecuencias para la metida de pata de Mario Delgado? (Más allá del ridículo).

@dhillonpost @natolisnuggets Still

@eltintero @natolisnuggets 21

Who’s got the plug to help with shipping around the country

Talked to a client this week who's rebuilding their entire system, Not because it's broken. Because their team refuses to use it. The tool technically works. Nobody wants to touch it. Here's what I keep seeing: - Companies optimize for features, not for the person clicking through it 50 times a day. - You can build the most powerful system in the world. If your team hates opening it, you've built an expensive paperweight. - Ttool adoption isn't a training problem. It's a "why did we choose this tool" problem. - Your CRM does everything—can your team find what they need in 3 clicks? - Your project tool has infinite custom fields—does anyone know what they mean? (this happened with our own previous CRM) I've watched more businesses struggle with their tools than with the work they couldn't do. The work gets done. The tool either helps or gets in the way.

@arturoangel20 Es un reverendo imbecile

Lo que faltaba. Mario Delgado celebra como un ¿¡logro!? que se haya revertido lo que él mismo anunció que se había modificado. Y dice que fue un "acto de justicia social". Ya casi se cuelga una medalla. O sea. Ya nos toman por estúpidos.

@SophieHuitzilin @mario_delgado @Claudiashein "la Sophie unineuronal"

@mario_delgado @Claudiashein Muy mal Siempre solapando la "conchudez" de los papás, deciden tener hijos y ya después no saben q hacer o a quién pedirle q se los cuide. Y claro, la escuela es su "guardería". Con ésto, solo se destapó los muchos q no planearon un hijo, y obvio eso con el tiempo "pesa".😒

El pasado 7 de mayo presentamos, por decisión unánime de todas las y los secretarios de educación del país, una propuesta de modificación al calendario escolar. Ante el llamado de nuestra presidenta, @Claudiashein, sobre escuchar y atender las propuestas de madres y padres de familia, cuidadoras, cuidadores, maestras, maestros y de toda la comunidad educativa. El calendario del ciclo escolar 2025-2026 se mantendrá sin cambios: la fecha de conclusión será el 15 de julio. Sin embargo, ante circunstancias extraordinarias o condiciones particulares de las entidades federativas, de conformidad con el artículo 87 de la Ley General de Educación, podrán realizarse ajustes al calendario escolar, siempre que se garantice el cumplimiento del plan y programas de estudio. Con este acuerdo que se llegó en la reunión del #CONAEDU, hay un acto de justicia social. Al sostener el calendario protegemos el derecho superior de las niñas, niños y adolescentes de México a la educación pública, transformadora y humanista que pondera la igualdad, la justicia y los saberes construidos en diálogo desde las comunidades. Este es el calendario vigente publicado en el Diario Oficial de la Federación el pasado 09 de junio de 2025.👇 calendarioescolar.sep.gob.mx/2025-2026

There's a difference between asking AI and delegating to AI. Asking: "Write me a follow-up email." Delegating: "Read yesterday's meeting notes. Check the proposal template in my Drive. Draft a follow-up that addresses Sarah's timeline concern from slide 4." The first gets you a template you'll rewrite. The second gets you something close to ready. Same tool. Completely different result. The variable is how much context you hand over before you start.

Had a call with a prospect today; decent budget, looking to replace their no-code stack. Their previous no-code agency: “We can’t build this anymore. You need full code.” This conversation wasn’t happening 12 months ago. The no-code era didn’t end with a bang. It’s ending with referrals out.

@vampipe @_JRLB_ uuu y al naco no le alcanza para ser un high-paying client yet

¿Se acuerdan de cuando @_JRLB_ fue fotografiado en Loro Piana? Ya tengo su historial de la base de datos de la empresa, €65 mil euros ha gastado. Y falta la de su esposa, se las comparto después. $1.3 millones de pesos, el bodoquito, en una sola tienda.

@zenorocha @marclou LOOOOVE resend, it's a fantastic, easy to use and worth-every-penny option

Resend started as a frustration. As an indie hacker, working nights and weekends, I felt like every email service was built for someone else. So I decided to create my own. Today @marclou runs 33 startups solo on it. I know Resend is not for everyone (and that's okay), but it feels nice to be able to build products for other indie hackers.

@arturoangel20 @mario_delgado ahora resulta q roba tiempo familiar... si los papas (y mamas) están trabajando!

Esto es lo que acaba de decir @mario_delgado en simple español. NO es broma: *Después del 15 de junio ya ni se enseña nada en las escuelas *Del 15 de junio al 15 de julio los papás usan la escuela de guardería. *La escuela no es guardería. *Maestros necesitan descanso.

@aglol0 🤮🤮🤮

No es por arruinar el Día de la Madre, pero creo necesario reconocer también a todas las mujeres que libre y responsablemente hemos elegido no ser mamás. Se habla poco de lo que esta lúcida elección aporta a la humanidad. Pero no estamos preparados para esta conversación.

@_javierivero @cursor_ai Nice! Felicidades

Big news: I’ve been selected as @cursor_ai Regional Lead for Mexico & Central America Calling all Mexican AI builders! I’ve got some credits if you wanna build with Cursor! Let’s build 🚀

@brandon @meow Really cool! Jealous on the office though 😉

What does @Meow do with its profits? Meow is profitable. People ask what we do with the profits. Four buckets: 1. Engineering and R&D. We're engineering-led. Bryce and I are both engineers and our founding team is early security engineers. The generational companies are all engineering-led. Most of the money goes into R&D and investing heavily in AI systems internally. 2. Licensing. We've set up a new subsidiary for the agentic future. The plan is to secure licensing so AI agents can send and receive dollars, USDC, USDT, SOL, ETH, BTC, and other currencies on rails we control. Security, compliance, and engineering is our bread and butter and vertically integrating this infrastructure is how we write the future, both from our direct business and as infrastructure. 3. Rate floors. We've continued to buy rate floors to hedge against a zero-rate environment. If rates ever go to zero, we make millions on the hedge. When we’re in one environment, we prepare for the exact opposite one, so the business is a beast in any environment. 4. Growth + a real office. We've been in WeWork the entire history of the company. After we hit eight figures in net revenue, we finally signed a four-year lease for a permanent NYC office. It'll let us scale the sales and engineering teams. I've always believed in growing within your profits and compounding. We'll get to $100M in net revenue faster than I imagined from keeping the team engineering-led, reinvesting profits wisely, and not taking on dilution or debt to get there.

@bentossell Oh I didn't catch that from anthropic

@eltintero anthropic staff tweeted about the tabs. and see codex. they’ll all be everything apps

did a breakdown in bensbites and sent out 20 mins ago - go read it if you’re interested 😁 will open source soon

The businesses that get the most out of AI aren't the most technical. They're the ones that took one week to understand how they actually operate before they touched a single tool. That week is the REAL work. Everything else is configuration.