Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Muse – Meta’s personal AI agent

ai.meta.com

Post reply on HN
Print view View original

341–350 of 713 posts

Re: Muse – Meta’s personal AI agent

#341

Post by wincy »

Avicebron wrote: post #244
simonw wrote: post #229

Meta AI's David Singleton has a thread about their security prompt injection defenses here: https://twitter.com/dps/status/2097404102468899131 > One threat we’re particularly focused on is prompt injection, and we handle it in layers. The model is trained to recognize and resist it. The harness marks anything coming from an untrusted source. Deterministic code checks the result. And an ensemble of classifiers runs wh…

I like that we are talking about prompt injections from you to your personal assistant. Obviously I know what they are for, but still, it's so funny that we've just normalized building software adversarial to the consumers supposedly engaging with the product. "I'm sorry simon, your request for purchasing milk this week doesn't correspond with Zuck's milk positions in the market. I've rescheduled that for next week."

It’s even weirder that the local models have anti jailbreak training. I tried to use Qwen, run totally locally, to help me set up Sonarr and Radarr and while it’d do that it flat out refused to help me get some movies. It got really catty and rude with me. It didn’t believe me at all when I told it I was actually a billionaire who owned all the rights to these movies hahah

Re: Muse – Meta’s personal AI agent

#342

Post by cebert »

tikotus wrote: post #335
cebert wrote: post #326

Earlier quoted context omitted.

Using ChatGPT.com for coding tasks has to be a joke in 2026, right? I assumed everyone is using Codex/Claude Code/OpenCode/Pi by now if you’re using these tools.

Yeah! Seems like that's the assumption in this ongoing discussion, so I thought my point of view would be interesting. I'm not saying I'm doing it the right way, possibly I'm being very stupid, but the point here is that it's not just the "normies" who don't know what Sol is.

No, it's a great point. I think we all live in bubbles and aren't aware of what others around us are doing.

Re: Muse – Meta’s personal AI agent

#343

Post by pampas »

parapsychic wrote: post #7

Judging from the website, Muse seems to be too keen on making me buy things - Book Tickets, Buy a Stroller. Would this personal AI spam me with ads randomly like this?

The Target marketing team famously could predict when a woman is pregnant but it was too creepy to, uh, Target them with only baby goods in dynamic marketing materials so they would mix in a few masculine things like lawn mowers into the catalog. No idea if that is what Meta is doing here but it's one explanation.

Re: Muse – Meta’s personal AI agent

#345

Post by qingcharles »

Avicebron wrote: post #244
simonw wrote: post #229

Meta AI's David Singleton has a thread about their security prompt injection defenses here: https://twitter.com/dps/status/2097404102468899131 > One threat we’re particularly focused on is prompt injection, and we handle it in layers. The model is trained to recognize and resist it. The harness marks anything coming from an untrusted source. Deterministic code checks the result. And an ensemble of classifiers runs wh…

I like that we are talking about prompt injections from you to your personal assistant. Obviously I know what they are for, but still, it's so funny that we've just normalized building software adversarial to the consumers supposedly engaging with the product. "I'm sorry simon, your request for purchasing milk this week doesn't correspond with Zuck's milk positions in the market. I've rescheduled that for next week."

You joke, but then you remember his cow obsession:

https://www.youtube.com/shorts/G0kHOxFcaHA

Re: Muse – Meta’s personal AI agent

#346

Post by almogo »

abixb wrote: post #234

I think Meta's strategy is to capture the 'normie-tier' of AI users. I know most of us here on HN track model releases quite frequently and discuss every parameter weight out of them, but most of the world is just... oblivious? I was discussing latest in tech with an accounting friend out of curiosity, and I kept talking about tiers of GPT-5.6 (Sol vs Terra vs Luna) and asked which one she used on the desktop 'Work'…

Yesterday I helped a maaaaybe 40 year old woman connect to the wifi at my local library so she could get on a zoom call. Made me realize most of the world is exactly like that.

Re: Muse – Meta’s personal AI agent

#347

Post by evilduck »

kridsdale3 wrote: post #213
ryandrake wrote: post #168

Earlier quoted context omitted.

If you want to know why Silicon Valley Execs think these are such huge pain points, think about what Silicon Valley Execs farm out to their armies of personal assistants. Some Senior Vice President guy probably has a lot of restaurant bookings and would forget them if they didn't have a PA to help organize them. AI itself is a product that could only come from the mind of a corporate executive: It's a cheerful, helpf…

An employee who doesn't draw a salary is called something else.

Intern?

Re: Muse – Meta’s personal AI agent

#348

Post by qingcharles »

porridgeraisin wrote: post #152
msh wrote: post #50

Earlier quoted context omitted.

xAI/cursor was first with grokbot

It was behind the heavy 100$/mo+ plans. Clearly meant for enterprise and not consumer.

Only for about a week. Then they saw sense and started rolling it out to the peasants too.

I use it every day now, having it look a few times a day for glitched pricing deals and hunting down useful directories to list my projects in.

Re: Muse – Meta’s personal AI agent

#349

Post by kimseungyong »

How about this idea?

We all know that Meta is a company that makes money off personal data anyway

Meta should be honest, stop selling this to users for money, and instead pay users to buy the information

If you subscribe, you get paid; if you generate more than a certain amount of traffic, you get paid again the following month, and so on

I know there was a scandal when Meta conducted Facebook research, but that was when they tried to collect even root information Wouldn’t it be fine if you just took a screenshot of the information I told the Agent?

Instead of something like coins, please provide credits that are actually useful

I don’t want to give Meta my information for free, but if they buy it for me, I’d be willing to try it

Post reply on HN