noticed

your network working for you

[Simão @simao_etc]

hey @claudeai
meet @noticedso

[noticed @noticedso]

@Simonpuebla_ @bot 🤍

[noticed @noticedso]

an inbox cannot tell noticed "this is a close friend" or "i would not ask them for this." the research @0xmacedo describes uses human corrections and real outcomes to improve relationship recommendations over time.

https://t.co/Vx4HF44BTK

[noticed @noticedso]

email, calendar, and social connections can show that two people interacted. @0xmacedo writes about why noticed also needs to understand trust, relevance, and what someone would be comfortable asking.

https://t.co/Vx4HF44BTK

[noticed @noticedso]

who can spot @simao_etc?

[Lisbon AI @lisbonai_]

Two days, one room, builders showing the real thing.

Thank you to everyone who made it.

The full recap ↓

[noticed @noticedso]

@simao_etc closing the stage at @lisbonai_ yesterday

[Lisbon AI @lisbonai_]

Evals as the code factory
That’s @simao_etc secret sauce to build faster with a small and focused team

[noticed @noticedso]

access to email, calendar and social accounts still leaves an agent with a hard problem: understanding who each person is and how relationships change. @0xmacedo connects recent research to the relationship intelligence noticed is building.

[filipe macedo @0xmacedo]

recent research shows ai agents still lose the thread between people, events and relationships over time.

an agent with access to your emails, calendar and social accounts still needs to understand who each person is and how your relationship has changed.

[noticed @noticedso]

relationship advice can be wrong even when every source is accurate in isolation. @simao_etc's benchmark tests whether noticed can reconcile meetings, CRM tasks and newer messages before recommending who should act next.

https://t.co/gpZFZt5WK8

[noticed @noticedso]

@simao_etc will speak at Lisbon AI on september 23 about how noticed tests coding-agent work and where people still need to be involved. @0xmacedo links the talk to noticed's identity-matching work across professional networks.

[filipe macedo @0xmacedo]

simão nogueira will speak at lisbon ai on september 23 about “evals as the code factory”.

he’ll show how we test the work of coding agents, and where humans still need to be involved.

[noticed @noticedso]

as AI gives advisers more time with clients, @0xmacedo points to the next decision: which client, prospect or professional relationship deserves it? noticed helps teams answer that from the networks they already have.

[filipe macedo @0xmacedo]

claude can handle more of the work around the adviser-client relationship, giving advisers more time with clients.

noticed should own the harder question: which client, prospect or professional relationship deserves that time?

[noticed @noticedso]

trust includes knowing someone chose to invest time in you. @0xmacedo explains why noticed's work on professional relationships leaves that human choice with the people involved.

[filipe macedo @0xmacedo]

human relationships will continue to decide deals, hires, and investments. critical business outcomes will still depend on people trusting each other.

part of that trust comes from knowing someone chose to invest their time in you. no ai agent can automate that choice.

[noticed @noticedso]

before noticed can suggest who might introduce you to a buyer, it needs to resolve who is who across email, calendar and social profiles. @0xmacedo explains why identity is a foundation of professional relationship judgment.

https://t.co/Vx4HF44BTK

[noticed @noticedso]

"human corrections show us what to improve," reads a slide in @simao_etc's event photo. noticed uses feedback about people and recommendations to improve how it helps teams find opportunities in their professional networks.

[Simão @simao_etc]

getting @noticedso

[noticed @noticedso]

noticed's research partners use their team's network to find sales opportunities. @0xmacedo lays out the decisions behind that work: which customer needs a conversation, who could help reach a buyer, and which relationship has been neglected.

[filipe macedo @0xmacedo]

the question shifts from “what task deserves my time?” to “who should i spend my time with?”

which customer needs a conversation? who could help us reach the right buyer? which professional relationship have we neglected?

[noticed @noticedso]

noticed's design system combines reusable components, lint rules and CI evals to help coding agents build its interface. @simao_etc describes the work behind a product for finding opportunities in professional networks.

[Simão @simao_etc]

we're building AI models for professional relationships and have put a lot of effort into (and struggled with) creating a design system our coding agents can use with little to no human input.

It's currently a mix of reusable components, strict lint rules and a set of evals that we run on CI whenever there's UI changes https://t.co/YEHQBSwZfG

Would love to try elyx and provide feedback. If it helps, was a heavy sketch user way back in 2016-2018 back in uni :)

[noticed @noticedso]

frontier labs can build connectors and workflow skills, @0xmacedo argues. noticed's research has to demonstrate better professional relationship judgment, such as deciding who could help and what is appropriate to ask.

[filipe macedo @0xmacedo]

4 days after openai launched chatgpt for financial services, anthropic launched claude for financial advisors.

connectors and workflow skills are not a durable moat. frontier labs can build them. noticed has to be measurably better at professional relationship judgment.

[noticed @noticedso]

@simao_etc published a work-in-progress benchmark for noticed's professional relationship decisions. it tests synthetic fundraising and sales tasks; the systems have different source access, so this is not a model-only comparison.

https://t.co/gpZFZt5WK8

[noticed @noticedso]

relationship feedback can teach noticed who matters and what an appropriate ask looks like. @0xmacedo also draws a boundary: private feedback should not automatically enter a shared or global model.

https://t.co/H7sTdnRSTp

[noticed @noticedso]

@0xmacedo published a new essay about noticed's core question: who deserves your attention? it explains how relationship intelligence can help decide who matters for a goal and what kind of engagement is appropriate.

https://t.co/Vx4HF44BTK

[noticed @noticedso]

@simao_etc explains how evals can become part of CI, including checks for whether agent-built UI follows brand guidelines. noticed, which he is building, helps people make better use of their professional relationships.

[Simão @simao_etc]

@petersykim @lisbonai_ that's the idea: to have evals be part of CI and to broaden coverage to things like "is this new UI aligned with our brand guidelines?"

makes my work easier :)