piet retweeted
Thank you Germany 🇩🇪 and UK 🇬🇧 for heroically sacrificing your economy to save the planet 💪
_
Lotomba, Deijl, St Juste, Borges, Nieuwkoop, Tengstedt, Kasanwirdjo, Ivanusec, Zerrouki en ga zo maar door.
The €3mn-12mn player who doesn’t play.
The most expensive transfer isn’t the €5M one.
It’s the €1M player who doesn’t play.
Because the real cost is hidden:
🔴wages over 2–3 years
🔴blocked squad spot
🔴pressure to fix the mistake
🔴lost time in development and results
One wrong decision creates three more.
1
151
piet retweeted
The ceo of Box says if you're not building for agents, you won't exist as a software company in 2-3 years.
@levie recalled how in the 90s, it was mainframes built for institutions. If you didn't build for the CIO or IT org, you didn't exist. You could build software for users, but no one cared.
Similarly, in the mid 2000s, it was software built for end users. If you didn't take that route, you got completely trounced. The likes of Yammer, Slack, Zoom, etc., ripped through enterprises with viral adoption. Bottom-up and completely user driven.
And now it's software built for agents. We're going to have trillions of agent runs. The systems that optimize for agents the most will win.
- lower latency
- better API exposure
- better system intelligence
- serves the right information when called
Agents will eventually choose their own tools on the fly and make purchasing decisions at runtime. All based on past experiences/online reviews/performance, sans any procurement process.
The more pragmatic version is companies swapping out bad tools / tech that don't work well with agents.
- crm doesn't expose the right APIs
- auth can't handle agent scale requests
- database has too much latency for agent workloads
Being good enough for humans won't work because they have to pass the litmus test for agents to make the cut.
De man @robdewijk is hallucinerend dan wel verkeert in bijzonder slecht welingelichte kringen.
Iraans kernwapen kan stabiliserend werken. bnr.nl/podcast/boekestijn-en… #BNR
1
13









