0:00

Today on the A.I. Daily Daily

0:02

Brief, a leaked internal memo

0:04

from Shopify that shows the

0:06

future of A.I. at the

0:09

workplace. Before that on the

0:11

headlines, cursor hits a million

0:13

users. The A.I. Daily Brief

0:16

is a daily podcast and

0:18

video about the most important

0:20

news and discussions in A.I.

0:23

To join the conversation, follow

0:25

the Discord link in

0:27

our show notes. We kick off

0:29

today with a story around the popularity

0:31

of AI coding platform cursor, which has

0:33

now officially hit a million users, honestly

0:36

without really trying. Platform creator AnySphere

0:38

has managed to catch lightning in a

0:40

bottle, with a three-year-old company not spending

0:42

a dollar on marketing, and yet still

0:44

becoming one of the fastest-growing startups ever.

0:46

In January, AnySphere reported a hundred million

0:49

on annual recurring revenue, and according

0:51

to Bloomberg sources, that figure had

0:53

doubled by March. The platform is now

0:55

seeing a million daily active active users.

0:57

Cursor currently offers two subscription tiers, a

0:59

$20 per month account for individuals and

1:02

a $40 business account. Any sphere

1:04

president Oscar Shulls mentioned that nearly

1:06

all of Cursor's revenue still comes

1:08

from individual users, many of whom

1:10

work at companies that aren't paying

1:12

for the tool directly. This suggests,

1:15

of course, that adoption is bottom-up,

1:17

with programmers using Cursor on their own

1:19

initiative because they want to improve their

1:21

productivity. This has been our experience

1:23

and how we see it happening, and how we see

1:26

it. At the same time, this year Shull

1:28

says that the company will begin actively

1:30

courting enterprise clients. They currently have 14,000

1:32

businesses signed up, despite a deliberately

1:34

obscure onboarding process. The company

1:36

now has hired their first salespeople and

1:38

say they had over 4,000 companies looking

1:40

to test the product in February alone. Any

1:43

sphere is currently in talks to raise

1:45

at a $10 billion valuation, which obviously

1:47

would help them expand from their just

1:49

60-person company right now. Next up a

1:51

little interesting nugget in the open AI

1:53

world the company is considering acquiring Sam

1:55

Altman and iPhone designer Johnny Ives AI

1:58

device startup the information reports that

2:00

open AI executives have discussed a $500

2:02

million acquisition price, but that a partnership

2:04

was also explored. Their sources say that

2:06

right now the company has only created

2:08

preliminary designs rather than full prototypes. Potential

2:11

designs include a phone without a screen

2:13

and other modalities for household devices. The

2:15

acquisition would be a very expensive aquahire and

2:17

would include the team of engineers currently working

2:19

on the device. Report suggested that

2:21

Altman is working closely with the startup but isn't

2:24

a founder and has an unknown stake in the

2:26

company. I think what gets people excited about

2:28

this is just the idea that the guy

2:30

who designed the iPhone might actually be fully

2:32

unleashed to think about the new user experience

2:34

of an AI-powered world. At the same time,

2:37

it does feel a little early for this sort

2:39

of acquisition, but who knows, Open AI

2:41

has a fat new valuation of 300 billion and

2:43

money to spend. Now, speaking of talent and

2:45

how much it costs to play in the AI game,

2:47

the battle for AI has now grown so intense

2:49

that Google would rather pay staff to do

2:51

nothing rather than jump ship ship to a

2:53

competitor to a competitor. According to

2:55

Business Insider, Google Deep Mind

2:58

has allegedly resorted to using

3:00

aggressive non-compete agreements. Some staff

3:02

are prevented from working for a

3:04

competitor for up to a year. The report

3:06

states that staff who want to leave have

3:08

been put on extended garden leave,

3:10

getting paid to wait out the duration

3:13

of their non-competes, getting paid to wait

3:15

out the duration of their non-competes. This

3:17

practice is very foreign to

3:19

the tech industry, with California's

3:22

ban on non-competes for residents.

3:24

Internationally, sectors like finance make extensive

3:26

use of non-competes to protect proprietary information

3:28

and prevent the loss of human capital.

3:30

Deep Mind has been able to make use of

3:33

non-competes as they're based in London, where the practice

3:35

is still legal. Commenting on the reporting, Google

3:37

said, our employment contracts are in line with

3:39

market standards. Given the sensitive nature of our

3:41

work, we use non-competes to protect our legitimate

3:43

interests. Former staffers say that six-month non-competes are

3:46

common for regular Deep Mind employees while

3:48

some regular Deep Mine employees, while some

3:50

senior researchers have year agreements. This is not

3:52

all that surprising, given that large amounts of churn

3:54

in top AI talent has been a hallmark of

3:56

the AI industry. Deep Mind itself suffered a ton

3:58

of brain drain in prior years. with co-founder

4:00

Mustafa Suleman leaving in 2022,

4:02

and Mistral co-founder Arthur Mench

4:05

departing in 2023. At the same time,

4:07

Google has also participated in their fair

4:09

share of talent poaching, aquahiring the character

4:12

AI team, including co-founder Nome Shizir,

4:14

last year. For Deep Mine's staff, looking

4:16

to jump ship, looking to jump ship,

4:18

looking to jump ship, these non-competes are

4:21

obviously a massive problem. One former

4:23

employee told business insider, who wants to

4:25

sign you for starting in a year.

4:28

That's, that's a year. The former employee said,

4:30

A.I. is interesting. It seems like the first

4:32

time in my career that you have this

4:34

insane race, like a space race. People really

4:36

feel like to be six months ahead a year

4:38

ahead, could make all the difference. Lastly

4:40

today, an update in this whole meta

4:42

situation from yesterday, the company has denied

4:45

rumors that the new Lamathor models

4:47

were specifically tuned to hit benchmarks.

4:49

Metas VP of generative AI Ahmed

4:51

Aldale posted on X. We're hearing some reports

4:53

of mixed quality across different services. Since we dropped the models

4:55

as soon as they were ready, we expect it'll take several

4:57

days for all of the public implementations to get dialed in.

4:59

We'll keep working through our bug fixes and onboarding partners. We've

5:02

also heard claims that we trained on onboarding partners. We've

5:04

also heard claims that we trained on test sets, that's

5:06

simply not true, and on-boarding partners. We've also heard claims

5:08

that we trained on test sets, that we trained on test sets, on

5:10

test sets, that we trained on test sets, on test sets, on test

5:12

sets, on test sets, on test sets, on test sets, on test sets,

5:14

on test sets, on test sets, on, on test sets, on test sets,

5:16

on test sets, on, on test sets, on test sets, on test sets,

5:18

on test sets, on, on test, on test, on

5:20

test, on test, on test, on test, on test,

5:22

on test, on test, on test, on test, on

5:24

test They said that company leadership suggested blending test

5:26

sets from various benchmarks during the post-training process, aiming

5:28

to meet the targets across various metrics and produce

5:30

a presentable result. Yesterday brought a whole new

5:33

wave of controversy as the AI community dug

5:35

into Lama Force performance. One of Metis Key

5:37

claims was ranking second on LM Arena. The

5:39

subjective benchmark where users vote for which

5:41

model produces the most pleasing output. LM

5:43

Arena has now released all 2000 head-to-to-head

5:45

battles involving Lama Fore for public review.

5:47

They noted that quote, Meda's interpretation of

5:49

our policy did not match what we

5:51

expected from model providers. Meda should

5:54

have made it clearer that Lamafor Maverick

5:56

0326 experimental was a customized model to

5:58

optimize for human preference. In essence, Alamarina

6:00

is calling out meta for deploying a

6:02

fine tune instead of a base model

6:05

for benchmarking. Although they did

6:07

also note that the practice wasn't technically

6:09

against the rules. AI developer Vic Kay

6:11

dug through the results giving one example

6:14

where Lama Ford defeated Claude 3.5

6:16

Sonnet. He commented, this is the

6:18

clearest evidence that no one should

6:20

take these rankings seriously. In this

6:22

example, it's super yappy and factually

6:24

inaccurate, and yet the user voted for

6:26

Lama 4. The rest aren't any better. or

6:29

if Ella Marina is infested with bots. Ultimately,

6:31

I think the big takeaway, and really the only one that

6:33

makes any sense, is to just put less emphasis

6:35

on the benchmarks. But then again, how are we

6:37

going to rate new things? I don't know. I don't

6:39

have the answers. For now, that's going to do it

6:41

for today's AIDily Brief headlines edition. Next up, the main

6:44

episode. A quick note, before we get into today's

6:46

ads. For those of you who are looking for an ad-free experience,

6:48

we are an ad-free experience, we are now, we are now for

6:50

an ad-free experience. We are now, we are now, we are now,

6:52

we are now, we are now, we are now, we are now,

6:54

we are now, we are now, we are now, we are now,

6:56

Right now the benefits of this are that you will

6:58

get the episodes without ads, and they will also

7:00

come out a little bit earlier. We'll be exploring

7:02

things like community and additional content in the weeks

7:04

to come, but for now I've heard from lots

7:06

of you that you are looking for an ad-free

7:09

experience, and so again it's patron.com/AI Daily Brief. Thanks

7:11

as always for supporting the show. Today's episode is

7:13

brought to you by Vanta. Trust isn't just earned,

7:15

it's demanded. Whether you're a startup founder navigating

7:17

navigating, navigating your first audit, or a

7:19

seasoned or navigating your first audit, or

7:21

a seasoned security professional, scaling, scaling your

7:24

GRC program, your GRC program, Proving your

7:26

commitment to security has never been more

7:28

critical or more complex. That's where Vanta

7:30

comes in. Businesses use Vanta to establish

7:33

trust by automating compliance needs across

7:35

over 35 frameworks like SOC 2

7:37

and ISO 2701. Centralized security workflows

7:39

complete questionnaires up to 5X faster

7:42

and proactively manage vendor risk. Vanta

7:44

can help you start or scale up your

7:46

security program by connecting you with

7:48

auditors and experts to conduct your

7:50

audit and set up your security

7:53

program quickly. Plus with automation and AI

7:55

throughout the platform, Vanta gives you time

7:57

back, so you can focus on building

7:59

your company. Join over 9,000 global

8:01

companies like Atlacian, Cora, and

8:04

Factory who use Vanta to

8:06

manage risk and prove security

8:08

in real-time. For a limited

8:10

time, this audience gets $1,000

8:13

off Vanta at vanta.com/NLW. That's

8:15

v-a-n-t-a.com slash-N-L-W for $1,000 off.

8:17

Hey listeners, are you tasked

8:20

with the safe deployment and

8:22

use of trustworthy AI? KPMG

8:24

has a first of its kind A.I.

8:27

Risk and Controls Guide,

8:29

which provides a structured

8:31

approach for organizations to

8:34

begin identifying A.I.

8:36

risks and design controls to

8:38

mitigate threats. What makes KPMG's

8:41

A.I. Risks and Controls Guide

8:43

different is that

8:45

it outlines practical control

8:48

considerations to help businesses

8:50

manage risks and accelerate

8:53

value. Today's episode is brought to

8:55

you by Plum. If you're building agentic

8:57

workflows for clients or colleagues, it's time

9:00

to take another look at Plum. Plum

9:02

is where AI experts create, deploy, manage,

9:04

and monetize complex automations. With features like

9:07

one-click updates that reach all your subscribers,

9:09

user-level variables for personalization, and the ability

9:11

to protect your prompts and workflow IP,

9:13

it's the best place to grow your

9:16

AI automation practice. Serve twice the clients

9:18

in half the time with Plum. Sign

9:20

up today at useplum.com, that's USE-P-L-U-M-B, dot

9:23

com,/NLW. Welcome back to the AI Daily

9:25

Brief. If you were anywhere around either

9:27

LinkedIn or X over the last 24 hours

9:30

or so, you probably have already seen

9:32

or at least seen conversation about this

9:34

internal memo from Shopify's CEO to buy

9:36

his lucky. Now this was apparently in

9:39

the process of being leaked when Toby

9:41

decided to just post the whole

9:43

thing so that it couldn't be

9:45

misappropri appropriated or misappropriated or

9:48

mis contextualized. What's capturing

9:50

people's attention with this

9:52

is that this is not a

9:55

gentle statement of suggestion when it

9:57

comes to AI This is a new mandate

9:59

from on high around how Shopify is going

10:01

to move to a reflexive AI posture

10:03

with everyone using AI in a fluent

10:05

and deeply integrated way. Capturing the most

10:07

attention, of course, is the provision that

10:10

says that Shopify folks are not allowed

10:12

to request new hires until they have

10:14

confirmed that AI can't do that job,

10:16

as this, of course, cuts right into

10:18

fears around AI-related job displacement, but the

10:20

story is much bigger than that. a

10:22

preview of the future, and a document

10:24

that, while seeming perhaps on the vanguard

10:26

or even to some extreme today, will

10:28

seem within a very short period of

10:30

time, probably about a year, completely obvious.

10:32

So let's actually read it first, and

10:34

then we'll dig into it. It's not

10:36

all that long. When he reposted it

10:38

to Twitter, Toby titled it, reflexive AI

10:40

usage is now a baseline expectation at

10:42

Shopify. He writes, team, we are entering

10:44

a time where more merchants and entrepreneurs

10:46

could be created than any other in

10:48

history. We often talk about bringing down

10:50

the complexity curve to allow more people

10:52

to choose this as a career. Each

10:54

step along the entrepreneurial path is rife

10:56

with decisions requiring skill judgment and knowledge.

10:58

Having AI alongside the journey and increasingly

11:00

doing not just the consultation, but also

11:02

doing the work for our merchants is

11:04

a mind-blowing step-function change here. Our task

11:06

at Shopify is to make our software

11:08

unquestionably the best canvas on which to

11:10

develop the best businesses of the future.

11:12

We do this by keeping everyone cutting

11:14

edge and bringing all the best tools

11:16

to bear, so our merchants can be

11:18

more successful than they themselves used to

11:20

imagine. For that, we need to be

11:22

absolutely ahead. Reflexive AI usage is now

11:24

a baseline expectation at Shopify. Maybe you

11:26

are already there and find a baseline

11:28

expectation at Shopify. Maybe you are already

11:30

there and find this memo puzzling. In

11:32

that case, you already use AI as

11:34

a thought partner, deep researcher, critic, or

11:36

pair or pair programmer for it. You've

11:39

heard me talk about AI in weekly

11:41

videos, podcast town halls, and summit. Last

11:43

summer I used agents to create my

11:45

talk and presented about that. I did

11:47

this as a call to action and

11:49

invitation for everyone to tinker with AI,

11:51

to dispel any skepticism or... that this

11:53

matters at all levels. Many of you

11:55

took up that call and all of

11:57

us who did have been an absolute

11:59

awe of the new capabilities and tools

12:01

that AI can deliver to augment our

12:03

skills, crafts, and fill in our gaps.

12:05

What we have learned so far is

12:07

that using AI well is a skill

12:09

that needs to be carefully learned by

12:11

using it a lot. It's just too

12:13

unlike everything else. The call to tinker

12:15

with it was the right one, but

12:17

it was too much of a suggestion.

12:19

This is what I want to change

12:21

here today. We have to change here

12:23

today. We are all lucky to work

12:25

with some amazing colleagues, the kind who

12:27

contribute 10X of what was previously thought

12:29

possible. It's my favorite thing about this

12:31

company. And what's even more amazing is

12:33

that for the first time, we see

12:35

the tools become 10X themselves. I've seen

12:37

many of these people approach implausible tasks,

12:39

ones we wouldn't have even chosen to

12:41

tackle before, with reflexive and brilliant usage

12:43

of AI to get 100X to work

12:45

done. In my on-leadership memo years ago,

12:47

I described Shopify as a red queen

12:49

race based on the Alice and Wonderland

12:51

race based on the Alice and Wonderland

12:53

story. In a company growing 20 to

12:55

40% every year you must improve by

12:57

at least that every year just to

12:59

re-qualify. This goes for me as well

13:01

as everyone else. This sounds daunting, but

13:03

given the nature of the tools, this

13:05

doesn't even sound terribly ambitious to me

13:08

anymore. It's also exactly the kind of

13:10

environment that our top performers tell us

13:12

they want. Learning together, surrounded by people

13:14

who are also on their own journey

13:16

of personal growth, and working on worthwhile,

13:18

meaningful and hard problems, is precisely the

13:20

environment Shopify was created to provide. deeply

13:22

connected to our core values of be

13:24

a constant learner and thrive on change.

13:26

These aren't just aspirational phrases. They're fundamental

13:28

expectations that come with being a part

13:30

of this world-class team. This is what

13:32

we founders wanted and this is what

13:34

we built. And this is what we

13:36

built. And so here we get to

13:38

the meat of it. What this means.

13:40

One, using AI effectively is now a

13:42

fundamental expectation of everyone at Shopify. It's

13:44

a tool of all trades today and

13:46

will only grow in importance. Frankly I

13:48

don't think it's You're welcome to try,

13:50

but I want to be honest, I

13:52

cannot see this working out today and

13:54

definitely not tomorrow. Stagnation is almost certain

13:56

and stagnation is slow motion failure. If

13:58

you're not climbing, you're sliding. 2. AI must be

14:00

a part of your GSD prototype phase.

14:02

The prototype phase of any GSD project

14:04

should be dominated by AI

14:07

exploration. Prototypes are meant for learning

14:09

and creating information. AI dramatically

14:11

accelerates this process. You can

14:13

learn to produce something other things.

14:16

AI dramatically accelerates this process. You

14:18

can learn to produce something other teams can

14:20

look at, use, and reason about in

14:22

a fraction of the time it used to take. 3.

14:24

We will add AI usage questions to

14:26

our performance and peer review review

14:28

questionnaire. Learning to prompt and load context is

14:31

important, and getting peers to provide feedback on how

14:33

this is going will be valuable. Four, learning is

14:35

self-directed, but share what you learned. You have access

14:37

to as much of the cutting-edge AI tools as

14:39

possible. There is chat. shopify. Shopify. Iow, which we've

14:41

had for Shopify. Iow, which we've had for years

14:43

now. Developers, Iow, which we've had for years now.

14:46

Developers have for years now. Developers have for

14:48

years now, Iow, which we've got for years now, Iow, Iow, I

14:50

know. I know. I know. I know which we have. I have. I have. I have. I

14:52

have. I have. I have. I have. I have. I have. I have. I have. I have.

14:54

I have. I have. I have. I have. I have. I have. I have. I have. I

14:56

have. I have. I have. I have. I have. I have. I have. Slack and

14:58

Vault have lots of places where people

15:00

share prompts that they developed. Five, before

15:02

asking for more headcount and resources, teams

15:04

must demonstrate why they cannot get what

15:06

they want done using AI. What would

15:08

this area look like if autonomous AI

15:10

agents were already part of the team? This

15:13

question can lead to really fun discussions

15:15

in projects. Six, everyone means everyone. This

15:17

applies to all of us, including me

15:19

and the executive team. Finally, he concludes

15:21

the path forward. AI will totally change Shopify

15:23

our work and the rest of our lives. We're all

15:26

in on this. I couldn't think of a better place

15:28

to be part of this truly unprecedented change than being

15:30

here. You don't just get a front row seat, but

15:32

are surrounded by a whole company learning and pushing things

15:34

forward. Our job is to figure out what entrepreneurship looks like

15:36

in a world where AI is universally available. And

15:38

I intend for us to do the best possible job of

15:40

that, and to do that, I need everyone's help. I already laid

15:42

out a lot of the AI projects and the themes

15:45

this year. Our roadmapemep is clear and our product will

15:47

better match our product will better match our

15:49

product will better match our mission. What we

15:51

need to succeed is our collective sum total

15:53

skill and ambition at applying our craft multiplied

15:55

by AI for the benefit of our merchants.

15:58

All right, so that is the piece like a... said,

16:00

a lot of the attention around it

16:02

has to do with that fifth bullet

16:04

about hiring. The Wall Street Journal piece

16:06

is titled Shopify says no new hires

16:08

unless A.I. can't do the job. Tech

16:10

Crunch's story is called Shopify CEO tells

16:12

teams to consider using AI before growing

16:14

headcount. And before we get into more

16:17

substantive conversation, we should note that some

16:19

people thought that this actually had to

16:21

do with what was going on in

16:23

the macro environment. Apple developer Nathan Lawrence,

16:25

for example, tweeted hard not to see

16:27

this announcement, as a which is probably

16:29

hitting many Shopify customers very hard. Others

16:31

chimed in that yes, maybe this was

16:33

a way to basically freeze hiring without

16:35

spooking Wall Street. And while I don't

16:38

think that there's nothing to this, obviously

16:40

every company is trying to adapt to

16:42

a very complicated and very tricky environment

16:44

that doesn't have a lot of answers

16:46

or clarity around how the future is

16:48

going to evolve, it strikes me that

16:50

to the extent that is true, its

16:52

main force is accelerating something which is

16:54

clearly already a priority for this company.

16:56

I think we don't have to dwell

16:58

too much on how much that macro

17:01

environment is playing into this particular policy

17:03

right now. To the extent that there

17:05

was concern or critique about this, it

17:07

usually fell into the category of either

17:09

one, this isn't the right approach to

17:11

executive leadership, basically forcing things by dictate

17:13

isn't likely to work. I will share

17:15

in a minute why I think that's

17:17

exactly opposite to true, but the other

17:19

side is of course just general concern

17:22

around what the implications are for labor.

17:24

And this wasn't even necessarily coming from

17:26

people who thought that Shopify was doing

17:28

anything wrong, but more of just a

17:30

general concern. Malcolm Peralty writes, the recession

17:32

we are already in and that will

17:34

worsen in large part thanks the trade

17:36

tariffs will massively increase the adoption of

17:38

AI and companies. The latest announcement from

17:40

Shopify's CEO about AI and headcount is

17:43

just the start. Companies will aggressively downsize

17:45

as an understanding of how to use

17:47

AI matures. No position is safe, no

17:49

level of talent is safe. The best

17:51

thing you can do is diversify your

17:53

skill set to diversify your skill set

17:55

in hopes it. Now I have a

17:57

slightly different read on this. First of

17:59

all, I think it's important here to

18:01

reintroduce this framework that I have of

18:03

Opportunity AI versus Efficiency AI. Efficiency AI

18:06

is a way of looking at our...

18:08

official intelligence, of course, as a tool

18:10

for efficiency, a way to get the

18:12

same outputs with fewer inputs. I've said

18:14

before frequently that I think that inevitably

18:16

some number of companies will view AI

18:18

strictly in this way. They will view

18:20

it as an exciting way to reduce

18:22

their costs, maybe get some plaudits from

18:24

Wall Street in the short term for

18:27

doing so, but ultimately be wildly out-competed

18:29

by the companies who instead view AI

18:31

as fundamentally opportunity generating. putting opportunity AI

18:33

in terms of labor, those who see

18:35

AI as a way to radically expand

18:37

the capability of their teams to deliver

18:39

new products, new services, better products, better

18:41

services, and whose focus instead of on

18:43

cost savings, on expansion and growth. Ultimately,

18:45

the companies that win or not the

18:47

companies that save the most are the

18:50

companies that grow the most. And I

18:52

think that even to the extent that

18:54

part of the immediate term motivation for

18:56

Shopify is bracing for recession. which could

18:58

more significantly impact them given the small

19:00

business entrepreneurs that they deal with and

19:02

who are their major customers, they are

19:04

still very clearly coming at this from

19:06

that opportunity AI growth AI type of

19:08

mindset. I mean, Toby articulates extremely clearly

19:11

here that the company has to grow

19:13

20 to 40% every year just to

19:15

stay still, and that because of that,

19:17

its people have to get 20 to

19:19

40% better year over year just to

19:21

stay still. Thinking in that way and

19:23

trying to exceed the stasis that comes

19:25

from base level growth is definitionally an

19:27

opportunity AI mindset. And I think it's

19:29

really important to note that number five,

19:32

the mandate to prove that AI can't

19:34

get done what you want to get

19:36

done before asking for more headcount does

19:38

not preclude more headcount. It just says

19:40

you have to start by seeing how

19:42

far AI can get you. To me,

19:44

this is as much about a mindset

19:46

shift around how people are thinking about

19:48

growth. as it is any sort of

19:50

soft hiring freeze. Let's put it this

19:52

way, this is exactly the same thing

19:55

that I expect inside Super Intelligent. We

19:57

stretch ourselves as far as we possibly

19:59

can, every single person in this company,

20:01

with AI, before we consider. are hiring

20:03

more people. And yes, of course, in

20:05

our startup world, that's because of resource

20:07

constraints, but with big enough ambition, everyone

20:09

is resource constrained. If you're trying to

20:11

double your revenue as a big public

20:13

company, even if you have all of

20:16

the resources in the world theoretically compared

20:18

to a small startup, you are still

20:20

definitionally resource constrained. And so I think

20:22

that this approach to hiring is just

20:24

going to seem like common sense in

20:26

the future. Now that's not to say

20:28

that there aren't some challenging implications here.

20:30

In fact, I think it's very likely

20:32

that this sort of becomes the default

20:34

type of quote-unquote job displacement that we

20:37

see in the short term from AI.

20:39

It's going to be about hiring freezes

20:41

and hiring slowdowns and people no longer

20:43

hiring junior versions of themselves than it

20:45

is about cutting out people who currently

20:47

exist and are employed. Because even if

20:49

you are in that growth mindset, There's

20:51

going to be a big transitional period

20:53

where you really want to see how

20:55

far your existing team can get superpowered

20:57

by these AI tools before you start

21:00

throwing more people at it. Now one

21:02

unanswered question that this brings up is

21:04

absolutely how we bring new people into

21:06

the workforce, what the future of mentorship

21:08

looks like, if no senior people who

21:10

do X ever hire junior people who

21:12

do X because they're just hiring agents

21:14

who do X, how are we going

21:16

to make that transition? Now I actually

21:18

have a ton of ideas. I'm fairly

21:21

optimistic here, but I think it's important

21:23

to recognize that there is theoretically a

21:25

challenge there that we will want to

21:27

address. In any case, this announcement was

21:29

about a lot more than headcount. In

21:31

fact, the thing that stood out most

21:33

to me was the mandate of it.

21:35

And this is, I think, the biggest

21:37

shift that you're going to see. For

21:39

a couple of years, there has been

21:41

a strong encouragement from many organizations to

21:44

start using and experimenting with AI. At

21:46

the same time, there have been a

21:48

couple of things holding organizations back. One

21:50

of them is limiting the tools that

21:52

are available to their people for concerns

21:54

around security, privacy, etc. Another is sort

21:56

of an illegitimacy of leadership. Basically, if

21:58

employees don't see their managers and executives...

22:00

actually making these transitions, they're less likely

22:02

to as well. A third piece of it is just

22:05

the normal inertia. Suggestions are only going to

22:07

take you so far, and that's why it both

22:09

doesn't surprise me that we are now seeing the

22:11

first mandates and why I think this is

22:13

exactly what's needed. Ultimately, people inside

22:15

big companies are going to adapt to

22:18

whatever they are evaluated on, and that's

22:20

why this third bullet is in some ways I

22:22

think the most significant of these. We will

22:24

add AI usage questions to our performance

22:26

and peer review review questionnaire. In

22:28

other words, you will be judged on whether you're

22:30

actually using AI or not. I don't think anything

22:33

short of that could actually prompt the

22:35

change that's going to be required. And

22:37

of course, bullet number six that this

22:39

applies to everyone is also related to

22:41

this same sort of leadership dynamic of

22:43

this. Ultimately, what matters is that there

22:45

is going to be accountability for this

22:48

AI transformation. Now one thing that I think

22:50

that Shopify is doing well here that other

22:52

organizations could take a hint from. is this

22:54

idea of learning being self-directed and being more

22:56

open and flexible to the tools that are

22:59

available to people. Like I said, another big constraint

23:01

on adoption is of course only having

23:03

some crappy old version of co-pilot that's

23:05

totally underpowered relative to all the things

23:07

you're using as an individual. That's why

23:09

we have so many people signing in with their

23:11

g-mails and using state-of-the-the-art tools instead of

23:13

bumbling around with whatever their work said

23:16

they could use. One of the single best

23:18

things that big organizations could do

23:20

is relax their defensive postures when

23:22

it comes to experimenting with new

23:24

tools, because otherwise these smaller and

23:26

mid-sized organizations are just absolutely going to

23:28

eat them for lunch. I did also want to

23:30

mention quickly this idea of AI being a part

23:32

of any sort of prototype or planning phase. This

23:34

is maybe the most specific injunction here,

23:36

and also one that we're seeing just

23:38

everywhere now. AI in general creates a

23:40

hugely differentiated opportunity to test out

23:42

more ideas and whittle them down

23:45

rather than trying to figure out

23:47

the best one before you've actually

23:49

run any tests. What's more the

23:51

rise of these vibe coding tools has

23:53

totally changed who has access to build

23:55

and prototype things. Already we're seeing inside

23:57

our own company and another start.

24:00

that we know, basically a ban on feature discussions

24:02

that don't involve actual working prototypes because there's

24:04

just no reason not to be able to

24:06

do that now. And in fact, going back

24:08

to the point just before about Enterprises, relaxing

24:10

their defensive posture, I actually think

24:12

that this is one area where

24:14

even very defensive organizations can move

24:16

the quickest. Sure, maybe lovable and bold

24:19

and these tools are not ready for

24:21

enterprise prime time when it comes to

24:23

actually integrating with big morass-like code bases

24:25

that are touched by thousands of people.

24:27

But when it comes to mocking up new features,

24:29

where you know that that mock-up is never meant

24:31

to see the light of day from an actual

24:33

code perspective, it's just meant to feed the speed

24:35

with which you can make new suggestions and have

24:37

new conversations about new opportunities, it

24:39

actually seems like something that perfectly

24:41

hits that low-stakes but high-impact Goldilocks

24:44

spot, I will be eagerly awaiting the first

24:46

of you corporate executives to email me and

24:48

tell me that your teams have embraced the

24:50

vibe prototyping life. So as I said at the

24:52

start, I believe that these Toby rules, or whatever

24:55

they end up being called, really do reflect where

24:57

the world is headed. I would encourage that

24:59

to the extent that these feel dramatic

25:01

now, the more dramatic they feel, probably

25:03

the farther you or your organization are

25:06

behind. Not every organization has to adopt

25:08

these exact same six mandates, but I

25:10

think that they should be thinking in

25:12

these sort of mandated type terms. And

25:14

again to reinforce, what's ultimately exciting about

25:17

this is that I do believe that

25:19

Shopify is coming at this with an

25:21

opportunity AI mindset. And my strong, strong

25:23

guess is that this makes the company

25:26

stronger, better performing, and ultimately is more

25:28

likely to grow headcount in the medium

25:30

and long term than if they didn't

25:33

have these mandates in place. Anyways guys,

25:35

that's going to do it for today's

25:37

a daily brief. Appreciate you listening as

25:39

always. Until next time, peace.