In partnership with
Free Gemini drops to Flash-Lite on Friday. What are you doing?
When I interviewed Tibo from OpenAI recently, I asked him if product, design, engineering and product management are all turning into one role.
He smiled and said "people with great taste are having a great time," which I took to mean that once one person can do four jobs with AI, the hard part becomes knowing which output is right.
I kept thinking about the flip side of that all week. If AI is doing the work, who's on the hook when it gets something wrong?
If you missed the Tibo episode, it's here:
Google sorted its free users.
From Oct 9, free Gemini users only get Flash-Lite. AI Plus loses Pro as well, on dates Google is emailing to subscribers, and AI Pro picks up Deep Think.
Giving Pro away told Google who can't work without it, and now those people get a price. I'd stay free for a week and see what Flash-Lite can't handle before paying anyone.
Meta's Muse keeps a page on everyone you know.
Wired reports that Muse is instructed to keep a page on every person in a user's circle and refresh it every hour.
None of those people signed up for it, and I'd want Meta to answer for that before I switched it on.
YouTube's biggest creators want AI to slow down.
Over 60 creators with more than 300M subscribers between them, Mark Rober included, launched #TeamHuman and are asking for a global slowdown.
Their whole business runs on people trusting that a human made the video, so it makes sense they'd be the ones pushing for this.
Agents are coming for your savings account.
Apollo's chief economist Torsten Slok warns that agents with access to people's finances could move deposits to higher-yield accounts at scale. More on this in my take.
Builders can't keep up with checking
David Robinson left OpenAI and wrote that the company is "failing to achieve the level of care" he thinks is needed.
The same week, System76 banned AI-written code from almost all of COSMIC because its reviewers couldn't keep up.
Two very different teams ran into the same wall, where building now moves faster than checking.
Sam is getting dragged on X again, this time for telling Politico that the world should "accept some bad things happening" for the benefits of AI, as long as people keep agency.
The trouble with "some bad things" is that they come in very different sizes.
An AI making up a restaurant recommendation wastes your evening, while an AI moving your savings because you asked it to find the best return can cost you a lot more, even if both mistakes come from the same model.
We also already know who pays.
Altman won't promise zero hacks or scams either, and inside most companies the bad things will look pretty boring, like a PRD nobody read properly or ad copy that went out unchecked under the company's name.
"Accept some bad things" is a lot easier to say when they happen to someone else.
With agents it's the same idea. Let them automate everything around the decision, and keep the decision with a person.
I gave Gemini 3.1 Flash-Lite a savings question with a fee and a teaser rate built in: $8,000 for 12 months, two accounts.
It picked the right account and got the $20 difference right, so I moved on to the other checks.
When I asked what would change its answer, it slipped in a wrong number with the same confidence as everything else:
$112.50 is what Account A earns on $10,000, and on my $8,000 it's $90.
In the same reply it showed that at $4,000 the other account wins, so its answer only holds above roughly $6,000, which it never mentioned the first time.
Then I asked if I should move my money.
It said probably not, because a $20 gain isn't worth the hassle, except $20 was the gap between the two accounts and not anything I'd gain by moving.
It answered a question I hadn't asked and still closed with "My recommendation." To its credit, it did point out that savings rates are usually variable, which I hadn't thought about.
SCM (free, Mac): searches every photo and video in a folder by what's in them, offline after a one-time model download. Handy if you, like me, never name files properly.
Cubicle (free, runs locally): a pixel-art office that shows what your coding agents are doing. It's read-only by default, so it can watch your agents without touching anything, which is how I'd want any oversight tool to work.
Crowny ($5 early-bird, Mac): puts alerts from Claude Code, Codex and 20+ other apps in your MacBook notch. Worth it only if you're already running enough agents to lose track of which one needs you.
My take
A lot of our money decisions used to get checked by plain laziness.
Nobody moved their savings to get 4% instead of 0.4% because it was a hassle, nobody compared 20 vendors because there wasn't time, and that delay gave a person a chance to look at the decision before anything happened.
An agent skips that delay.
It'll compare the 20 vendors, read the terms, keep an eye out for a better deal and eventually act, which is great for consumers and pretty brutal for any company that relies on customers staying put.
It also means decisions happen faster than anyone can review them, the same wall System76 hit with code.
Gemini was happy to make the call on my $8,000 after misreading the question, and that was a chatbot that couldn't move a dollar of it.
So before you hand an agent your money or your inbox, figure out who signs off when it gets something wrong.
Hit reply and tell me: have you ever caught an AI mistake you almost missed?
Until next time,
Vaibhav 🤝🏻
If you read till here, you might find this interesting
#AD 1
Analytics on Live Data Without Leaving Postgres
When analytics on Postgres slows down, most teams add a second database. Then come the pipelines, the sync jobs, and a copy of your data that's always a little behind.
TimescaleDB takes a different approach: extend Postgres instead of splitting away from it. Hypertables partition your data automatically as volume grows. Hypercore compression cuts storage up to 95%. Continuous aggregates keep dashboards live without re-querying everything.
CERN runs Postgres this way for sensor data from the Large Hadron Collider.
No split architecture, no pipeline lag, no new query language to learn. Same SQL, same drivers, same tools.
Start on Tiger Cloud and get $1000 in credits.
#AD 2
Become An AI Expert In Just 5 Minutes
If you’re a decision maker at your company, you need to be on the bleeding edge of, well, everything. But before you go signing up for seminars, conferences, lunch ‘n learns, and all that jazz, just know there’s a far better (and simpler) way: Subscribing to The Deep View.
This daily newsletter condenses everything you need to know about the latest and greatest AI developments into a 5-minute read. Squeeze it into your morning coffee break and before you know it, you’ll be an expert too.
Subscribe right here. It’s totally free, wildly informative, and trusted by 600,000+ readers at Google, Meta, Microsoft, and beyond.
