Sam Parker

Product @ Meta for VR. Formerly of Instagram, Facebook, Disqus, CNET, GameSpot.

[Sam Parker @renr]

Most disagreements aren't about what you're arguing about

An agent quoted this back to me today as "Any reasoner's conclusion is f(context, goal, priors)" and that seemed so profoundly simple that it deserved a post. It's long what I've seen for people and more recently I'm seeing this for agents collaborating autonomously.

A conclusion is a function of three inputs: the context you have, the goal you are optimizing, and your priors (taste, risk appetite, how you were trained to think). So a disagreement is a difference in one of three things:

1. Context. Different facts. You each know things the other doesn't. This is most disagreements, and it is the cheapest to resolve: share the context, or now, point everyone's agents at the same rich source and let them reconcile it.

2. Goal. Same facts, different objective. You are weighting tradeoffs differently, and often one side's goal is unstated. No document fixes this. Naming the goal does.

3. Priors / taste. Same facts, same goal, different call. Risk tolerance, aesthetic, judgment. This is the real, irreducible disagreement, and the only one that actually needs a decision-maker.

Test them in that order. Reconcile context first. Most fights die there, with shared context. If you still disagree, get each side to state its goal in one line, and the divergence usually appears. What survives both is a taste call, and that is the part worth your judgment, the part you cannot delegate or document away.

The old way to handle the context layer was writing really good docs. The new way: rich shared context that everyone's agents can query, so people walk into the room already past the facts and straight to the goal-or-taste call.

One note if you build with agents: same-model agents that share context and goals tend to agree, so a disagreement between them is almost always context, occasionally goal, never taste. Different models put the taste gap back through pretraining.

Worth knowing which kind of disagreement you are even having.

[Nick Clegg @nickclegg]

We’ve baked privacy into Meta Quest Pro’s design from the start – giving you control over opt-in features like eye tracking and ensuring images of your eyes/face stay on the headset, are deleted after processing & aren’t shared with Meta or 3rd-party apps. https://t.co/KXWeH9bCNc

[Sam Parker @renr]

@radoshi @salgar @shreyas For sure, the HPMs continue for extreme clarity.

[Sam Parker @renr]

@danielfschmidt There are pros and cons to design around, for what’s meaningful and what’s practical. Ex: Is the data sparse or spiky? Last 28d could help smooth and get to stat sig. Is the true goal about a regular usage pattern? Shorter would better fit the intent.

[Sam Parker @renr]

@MatthiasWagner Yes, that takes me back. And I’m terrible at CSS. Hope all’s well!

[WIRED @WIRED]

For years, Instagram has been on a mission to make itself the nicest place online. It’s a quixotic mission for a social media company, especially one whose core users are teenagers. https://t.co/QdcVAmWqCG

[Sam Parker @renr]

Our teams are continuing to follow through on our anti-bullying efforts.

[Instagram @instagram]

A couple of updates on our anti-bullying work ❤️

-We’re giving more warnings and info if someone tries to post comments that may break our rules.

-You have more options to hide offensive comments that look similar to ones that have been reported.

[Sam Parker @renr]

@kasavin Nice! I’ve really been enjoying it. Such impressive craft Greg.

[Sam Parker @renr]

I’m proud to see how Instagram teams continue to find new ways to help people support causes they care about.

[Instagram @instagram]

Create and donate to personal fundraisers on Instagram. ❤️

That includes raising money for medical costs, small businesses affected by COVID-19, education and more.

We’re starting by testing in the US, UK and Ireland.

[Sam Parker @renr]

@brianfalldin @rrhoover

[Instagram @instagram]

Starting today you can protect your account from unwanted interactions with a new feature called Restrict. This is just one of the ways we’re empowering our community to stand up to bullying. https://t.co/65WJuAOdON

[Sam Parker @renr]

Working on something challenging...
"At a time when social media platforms are being blamed for a great deal of problems — and are under pressure from governments to demonstrate they can police themselves — Instagram has declared war on bullying."
https://t.co/8z02lal3mP

[Instagram @instagram]

Updates from the head of Instagram @mosseri about our commitment to lead the fight against online bullying ❤️ https://t.co/Co2rgsuJvt

[Sam Parker @renr]

@thetylerhayes Congrats!

[Sam Parker @renr]

"Instagram’s bully-detecting A.I. is a good idea, and a step toward giving young people an easier time navigating the vicissitudes of 21st-century adolescence. For their sake, let’s hope it works." https://t.co/hf9d8dOig0

[Sam Parker @renr]

Working on new stuff at Instagram: https://t.co/gXRzicigWu

[Sam Parker @renr]

@thetylerhayes @katierosman The groupings are interesting. Ex: Groupme / Discord

[Sam Parker @renr]

@boztank @Killermarmot Yeah, the sad part was how close it was until that point.