@luchoalfonso_ Que Bravilo saliera de una herramienta interna de Jobly pasa más de lo que se cuenta. Muchas veces el producto ya está escondido en el negocio que uno quiere cerrar. ¿Cuál fue la señal de que clientes de afuera sí lo iban a pagar?
@juancpedreira La parte difícil va a ser la misma de Google My Business: mantener la información al día. Si el agente responde con precios o horarios viejos, el cliente culpa al negocio, no a ChatGPT.
@frontendmentor The design engineer path looks like the safer bet. Agents can write a decent component, but they still can't tell you why a form feels wrong to the person filling it in.
@BHolmesDev Same rule you'd give a junior: run the checks before asking for a person's time. The PR should show what the agent already caught and fixed, not a wall of its own comments.
@gimenete The cheap version of that is boundaries an agent can't cross by accident. Clear module edges and tests that fail loudly do more than any prompt asking it to be careful.
Canada's Real-Time Rail went live this week after 11 years, with eight launch members and none of the Big Six banks. Wealthsimple and Koho are on it from day one. Shipping with the ones who are ready was the right call. Now the hard part is getting everyone else on.
Ramp says AI token spend grew only 1.5% across August and September after a 42% jump in July, while usage kept rising. Apple reportedly cut engineers' daily AI budgets from $300 to $200 in October. Per-seat AI budgets now belong in the delivery plan from day one.
@PovilasKorop I don't think review goes away. It moves up a level, from reading every line to owning the tests, the limits and the final sign off. The open question is who checks the agent that does the reviewing.
@addyosmani The handoff part is where most teams fail first. A one line ticket with the real expectations in someone's head breaks a new hire and an agent the same way.
@levie The limit moves from doing the work to checking it. When ten agents finish in an hour, someone still has to decide what ships, and most teams haven't planned for that review load.
Wealthsimple's Penny prepares transfers and trades, then waits for the client to confirm. Planned ChatGPT and Claude access works the same way: the client sets the limits. Lab829 published that pattern in March. The agent drafts. The owner sends.
Bancolombia dejó que Nequi, creado dentro del banco, compitiera por sus propios clientes. Hoy suma 24 millones y tiene licencia propia. Fidel Vargas, su CIO, explica la regla: empezar adentro y separar solo cuando el cliente lo exige. #TechLATAM
How would you rather pay for an ePub reader you use every day? One price up front, a yearly subscription, or free with a paid upgrade? And what would make that price feel fair to you?
Lab829 asking, for research on an ePub reading app for iOS. No link.
If you used an iOS ePub app that's gone now, like Marvin 3, KyBook 3 or Stanza: what is the one thing you still haven't found anywhere else?
Lab829, doing reader research. No link.
More than 40% of pull requests on GitHub now start with an agent's first draft. In March it was 17%. This week GitHub made agentic workflows generally available inside Actions. Writing code is no longer the constraint. Review capacity and ownership are leadership decisions now.
A sketch worth keeping doesn't have to stay in the browser. Download Project in CodeRunner hands over the full Vite app, ready for your own editor.
Unzip it and open it in your own editor.
Free in the browser. Link in the reply.