Victor
banner
panvictor.bsky.social
Victor
@panvictor.bsky.social
43 followers 33 following 240 posts
Product SEO @HubSpot. Ex-WPP Search Director. Son of immigrants. TKK Fried chicken adjacent. Rooting for a kinder, gentler world for my girls. He/Him/先生 #A11y #ASD
Posts Media Videos Starter Packs
Google won over schools with Chromium.

So... ChatGPT Atlas being hyped up by their Education team... means... they're going after students?

That will be interesting if true.
Somewhere in the range of 10~20% of EU MAU also use LLMs wouldn't be too crazy. There's overlap.

It does make you question how good they are at bot-detection and deduping accounts.

searchengineland.com/chatgpt-sear...
ChatGPT search has 41 million average monthly users in EU
While ChatGPT Search grew from 11.2 million to 41.3 million "recipients," it is still dwarfed by Google Search.
searchengineland.com
youtu.be/6PQB1S5sZQ0?...

Shared while my mind is thinking about ChatGPT Atlas and "assistants" in general.

Context: I grew up with maids serving the family. I also did a whole paper while in AP US history about indentured servants.
Parts Unknown : Singapore
YouTube video by All About Food
youtu.be
Oh, and bonus:

Your prompts should be an index. An arbitrary collection of prompts that you believe reflect the health of your brand or industry.

Indexes should be updated and those changes communicated.

Indexes are comprised of different things arbitrarily, but very intentionally.
I think the bigger fight we have right now is prompt volumes since uh, clustering sizes are an arbitrary decision.

Without getting into how panel data is collected, I'd rather focus on real-user metrics and experiences and focus on brand and influencing those moments of decision-making.

/End
In other words, I agree with the Waikey/Dan Petrovic method as well - and frankly worry that others are copy+pasting queries from GSC, overinflating bot activity.

Thinking back, we swapped the order of goals 3-4 with 5-6, as they were closest to ROI.

3/N
Goal 3: Your entity is mentioned in topics you care about grounded
Goal 4: Your entity is mentioned in topics you care about ungrounded
Goal 5: Your entity recommended in decision making grounded
Goal 6: Your entity is recommended in decision making ungrounded

2/N
Generally agree.

1. Grounded vs ungrounded.
2. Ungrounded and brands are in the answer
3. Grounded and brands are in the answer.
4. Some prompts where brands are not in the answer.

Goal 1: Your entity is accurately portrayed ungrounded
Goal 2: Your entity is accurately portrayed grounded

1/N
👀https://www.uschamber.com/co/about-us/authors
I'm very glad that the ?hints=search&q= hack to force QFO doesn't directly jailbreak the protections built into ChatGPT.
Recent-ish skills acquired/re-acquainted to:

Playing chess
Solving a Rubik's cube
Juggling a soccer ball 5x times
Overhead tennis serves
Double stop exercises on the violin
Being a piano accompaniment

Because of my kids.

At least two of these I actually enjoy.
This is the most gutsy thing I've seen all day.

kayak.ai

Check out the robots.txt and sitemap to get a flavor of what they think the premise is how the web will evolve.
TIL AI Psychosis

That switch from 4o to 5 has that on full display.
Feel the inflation > feel the AGI
IMO, live the simulation to the fullest of transparency, joy, and good vibes over secrecy, fear, and anxious vibes.

As always, it's the friends we made along the way.
If AI can create simulations, what's to say we're not in the matrix in a simulation, or that we don't become capable of creating simulations where there are entities unaware that they are in a simulation?

An old question revisited in the age of AI.
ADAPT OR BE REPLACED BY SOMEONE ELSE WHO HAS EVOLVED WITH AI

Me: Chill with the caps. It's a good idea to keep a few colleagues whose role is to play devil's advocate because group think is dangerous.

Actually, you're right. I give up.

Let them create a devil's advocate agent.
4. He taught himself divination from ancient texts. All of his grandchildren's destinies were determined upon birth and name. Our names all sound terribly non-standard.
Some (tall?) tales:
1. He was top of his class - selected to go to Japan to further serve the Japanese empire.
2. He'd swim between Kinmen and Xiamen to co-ordinate trade deals. Swim.
3. He turned knowledge into wealth, and wealth to charity. A big donor to a Mazu temple.
Today is 10/10🎆

1. It marks the uprising against the Ching Dynasty, which ended in 1911

2. It's the national day celebrated in Taiwan, the sovereignty at the core the AI boom cause TSMC makes your AI chips

3. The fireworks celebration on Taipei 101 always looks like a terrorist attack aftermath
Oddly satisfying.
MIT Economists Amy Finkelstein and Benjamin Olken with their matching "Supply" and "Demand" numberplates.
#econsky
Some fun questions:

1. Can we detect LLM agents and goad them to site search when they hit a 404?

2. Can we detect existing users and provide "this is where you last left off?"

3. Can we double-down and just ask... "most people complete X or Y after where you left off?"
If we serve a 404 page that actually is a 404 http status code...

But we infer a confidence score using personalization and cookies... to implement a meta refresh..

That then gets you to a 200 status page with noindex on it...

I think we could spin up a better 404 user-experience.

Am I crazy?
On my father's deathbed, the tale of how a family of "communist" spies were provided safe haven during the 228 incident unraveled.

To my grandfather, they were neighbors first over political thought. To my father, it was a memory of my grandfather's greatness.

A greatness he chased his whole life.
Almost more year down, 3 more to go...

I wish politicians were more boring and less of a reality show.

This attack on free speech is concerningly creeping towards things you experience in autocratic regimes.