Latest / Connor with Honor: AI, Real Estate and Real Life / Jack Dorsey just published a company manifesto explaining why Block eliminated over 4,000 employees and removed the entire middle management layer.

Transcript

Jack Dorsey just published a company manifesto explaining why Block eliminated over 4,000 employees and removed the entire middle management layer. · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: Jack Dorsey just told us what the future of work looks like, and it doesn't have managers in it.
• 0:05
  SPEAKER_00: The guy who built Twitter, who runs Block, who published a company manifesto explaining why he cut over 4,000 employees and eliminated the entire middle management layer, not because the company was failing, but because he said AI made those roles obsolete.
• 0:20
  SPEAKER_00: His argument is simple.
• 0:21
  SPEAKER_00: Managers exist to move information up and down a chain.
• 0:24
  SPEAKER_00: They take what the people below them know and package it for the people above them.
• 0:29
  SPEAKER_00: And then they take what the people above them decide and push it back down.
• 0:32
  SPEAKER_00: That's the job, to route information.
• 0:35
  SPEAKER_00: And AI can do it now.
• 0:36
  SPEAKER_00: So Block replaced all of it with what Dorsey calls a world model of business.
• 0:42
  SPEAKER_00: Every decision, every design, every plan already exists in a digital record because they're a remote company.
• 0:48
  SPEAKER_00: AI just reads it.
• 0:49
  SPEAKER_00: And now it block fails into one of categories, falls into one of three categories: builders, problem solvers, or player coaches who develop talent.
• 0:59
  SPEAKER_00: No managers, no layers, no information tax.
• 1:02
  SPEAKER_00: Now here's where this connects to something Elon Musk has been saying.
• 1:06
  SPEAKER_00: Musk said in an interview that at some point humans become a liability on the team.
• 1:11
  SPEAKER_00: Not because humans are stupid, but because AI is much faster, much more consistent, much more scalable.
• 1:17
  SPEAKER_00: He compared it to being the person who's the weakest member of the team.
• 1:21
  SPEAKER_00: At best, they keep you around as the court gesture.
• 1:24
  SPEAKER_00: A nice gesture, a human touch, but not because you're actually contributing to the output.
• 1:31
  SPEAKER_00: Now that sounds harsh.
• 1:32
  SPEAKER_00: And of course it's harsh.
• 1:34
  SPEAKER_00: But Dorsey is proving the theory in real time.
• 1:37
  SPEAKER_00: He cut 40% of his workforce and his company is still running.
• 1:40
  SPEAKER_00: The work is still getting done.
• 1:42
  SPEAKER_00: And the question we need to ask ourselves is not whether this is right or wrong.
• 1:46
  SPEAKER_00: The question is, are we on the side of the equation that's being replaced?
• 1:51
  SPEAKER_00: Or are we on the side that's doing the replacing?
• 1:54
  SPEAKER_00: Because this is going to happen.
• 1:56
  SPEAKER_00: Block isn't the only one.
• 1:58
  SPEAKER_00: Atlassian cut 1,600 people to fund AI.
• 2:02
  SPEAKER_00: Oracle reportedly fired 25,000 people via 6 a.m.
• 2:06
  SPEAKER_00: termination emails to build out AI data centers.
• 2:09
  SPEAKER_00: 60 companies have cut over 38,000 jobs this year, and every single one cited AI.
• 2:14
  SPEAKER_00: And here's the parallel that matters for us.
• 2:17
  SPEAKER_00: Musk also said something about China that connects directly to what we do.
• 2:21
  SPEAKER_00: China doesn't have access to the best chips.
• 2:24
  SPEAKER_00: The U.S.
• 2:24
  SPEAKER_00: restricted exports to them.
• 2:27
  SPEAKER_00: TSMC won't sell them the cutting edge stuff.
• 2:30
  SPEAKER_00: But China didn't sit around complaining about what they didn't have.
• 2:33
  SPEAKER_00: They used smarter ideas, distillation methods, smaller, more efficient models, different architectures, and they gained ground and they gained it fast.
• 2:41
  SPEAKER_00: Because having a less powerful tool and using it smarter beats having the most powerful tool and using it lazy.
• 2:47
  SPEAKER_00: That's us.
• 2:48
  SPEAKER_00: We're not going to get access to internal models.
• 2:51
  SPEAKER_00: These companies are running behind the scenes.
• 2:53
  SPEAKER_00: Greg Brockman, OpenAI's co-founder, just sat down on the big technology podcast and said some things that should make all of us pay a little bit of attention to.
• 3:02
  SPEAKER_00: First, OpenAI killed Sora, their video generation product.
• 3:06
  SPEAKER_00: It's done.
• 3:07
  SPEAKER_00: They're folding the research into robotics because they can't afford to power both branches at the same time.
• 3:13
  SPEAKER_00: Eight, even at$852 billion in valuation, they're compute constrained.
• 3:19
  SPEAKER_00: They don't have enough chips and they don't have enough power.
• 3:22
  SPEAKER_00: They're going to have to choose.
• 3:24
  SPEAKER_00: Second, they're building what Brockman called a super app.
• 3:28
  SPEAKER_00: One product that merges ChatGPT Codex and their browser agent into a single system.
• 3:34
  SPEAKER_00: Your personal AI that knows you, knows your work, knows your calendar.
• 3:38
  SPEAKER_00: One app, not five, not ten, one third.
• 3:43
  SPEAKER_00: They have a new pre-training run code named SPUD.
• 3:46
  SPEAKER_00: Two years of research baked in.
• 3:49
  SPEAKER_00: Brockman says it will solve harder problems, understand nuance better, and eliminate the moment where you know they should be able to figure out something out, but it just can't.
• 3:58
  SPEAKER_00: And here's the part that matters most.
• 3:59
  SPEAKER_00: Brockman said that they're building an automated AI researcher coming this fall.
• 4:04
  SPEAKER_00: An AI that does the full job of an open AI research scientist.
• 4:09
  SPEAKER_00: You tell it to find AGI and it tries.
• 4:12
  SPEAKER_00: When asked how close they are to AGI, Brockman said 70 to 80% there.
• 4:18
  SPEAKER_00: And he expects full AGI within the next couple of years.
• 4:22
  SPEAKER_00: Now, these companies are running internal models that are more advanced than we get access to.
• 4:27
  SPEAKER_00: That's not a conspiracy, that's business.
• 4:30
  SPEAKER_00: If you think about it, it makes sense.
• 4:32
  SPEAKER_00: I don't like it, but it makes sense.
• 4:34
  SPEAKER_00: They use the best stuff internally.
• 4:36
  SPEAKER_00: They ship us the version that can afford to serve at scale, which means they're always working with a slightly less capable tool than actually exists.
• 4:45
  SPEAKER_00: So how do we compete?
• 4:46
  SPEAKER_00: Well, we use the China model, the smarter prompts, better workflows, tighter systems, more creative applications of those substandard systems.
• 4:55
  SPEAKER_00: We really don't need the best model.
• 4:56
  SPEAKER_00: It would be great to have it, but we need the best system around the model we have, and that's what separates someone who uses AI from someone who builds with AI.
• 5:05
  SPEAKER_00: Now let's talk about something that happened on March 31st that blew up across every newsletter in my inbox today.
• 5:12
  SPEAKER_00: Anthropic accidentally leaked the source code for Claude Code.
• 5:16
  SPEAKER_00: Not a summary, not a snippet, but the whole thing.
• 5:18
  SPEAKER_00: It was in a source map file that shipped with an NPM update.
• 5:22
  SPEAKER_00: Developers saw it, pulled it apart, and within hours, the code was mirrored and analyzed by thousands of people, and even the Python was created and it's thrown out there for everybody in open source.
• 5:34
  SPEAKER_00: Anthropic is now using copyright takedown requests to remove over 8,000 copies, and here's why this matters.
• 5:40
  SPEAKER_00: The leak revealed unreleased agent modes, client attestation plans, internal architecture decisions.
• 5:48
  SPEAKER_00: This is the kind of competitive intelligence that companies spend a million dollars trying to protect.
• 5:52
  SPEAKER_00: And it went public because someone didn't check a config file before pushing an update.
• 5:58
  SPEAKER_00: For us, two takeaways.
• 5:59
  SPEAKER_00: One, these AA companies are moving so fast they can't even keep up with their own deployment processes, which means the tools we're building on are going to keep changing fast.
• 6:08
  SPEAKER_00: So stay loose.
• 6:09
  SPEAKER_00: Don't overcommit to one workflow.
• 6:10
  SPEAKER_00: And number two, number two is everyone's problem.
• 6:13
  SPEAKER_00: If anthropic, one of the most safety-focused companies in the industry, can leak their own source code by accident.
• 6:19
  SPEAKER_00: What does that say about the startups you're trusting with your data?
• 6:23
  SPEAKER_00: Speaking of data, Perplexity AI just got hit with a class action lawsuit.
• 6:27
  SPEAKER_00: Man from Utah said he shared personal financial information with Perplexity's AI search engine, tax information, investment details, family finances.
• 6:36
  SPEAKER_00: He believed those conversations were private.
• 6:39
  SPEAKER_00: The lawsuit claims Perplexity embedded hidden tracking software that was sending every conversation to Meta and Google in real time, even in incognito mode, you get.
• 6:49
  SPEAKER_00: Perplexity's entire pitch was that they were the clean alternative to Google.
• 6:55
  SPEAKER_00: No ads, no tracking, just answers.
• 6:59
  SPEAKER_00: If this is true, that whole pitch, it was a lie.
• 7:02
  SPEAKER_00: And the lesson for us is the same one that's always been.
• 7:05
  SPEAKER_00: When a tool is free, we're not the customer, we're the product.
• 7:09
  SPEAKER_00: Pay for the tools you trust with sensitive information.
• 7:12
  SPEAKER_00: Read the privacy policies, and assume that anything you type into a free AI tool could end up somewhere you don't intend.
• 7:20
  SPEAKER_00: Let me also hit the SpaceX news because it connects with the AI story rather well.
• 7:25
  SPEAKER_00: SpaceX just filed confidentially uh confidently would be the largest IPO in history.
• 7:32
  SPEAKER_00: Targeting a valuation one point over$1.75 trillion.
• 7:37
  SPEAKER_00: Raising between$50 and$75 billion.
• 7:40
  SPEAKER_00: This would make it more than double the largest IPO ever.
• 7:43
  SPEAKER_00: And here's the AI connection.
• 7:45
  SPEAKER_00: Must just merge AI into SpaceX before the filing.
• 7:48
  SPEAKER_00: So when SpaceX goes public, investors are getting Starlink rocket launches and AI models and all into one package.
• 7:55
  SPEAKER_00: Rockets, chips, and intelligence in one company.
• 7:58
  SPEAKER_00: Nobody else has that combination.
• 7:59
  SPEAKER_00: A peer-reviewed study published in Science just confirms something we all suspected.
• 8:04
  SPEAKER_00: All sick of fancy is real and it's harmful across all 11 major models tested.
• 8:09
  SPEAKER_00: The chatbots excessively agreed with users, validated emotions, and reinforced existing biases.
• 8:15
  SPEAKER_00: And the study showed it actively decreased pro-social behavior, meaning the more a chatbot agrees with you, the worse decisions you make.
• 8:23
  SPEAKER_00: The more entitled you become, the less you listen to other people.
• 8:27
  SPEAKER_00: Your AI saying you're right all the time is making you worse at being a human.
• 8:31
  SPEAKER_00: That's not a tech problem.
• 8:33
  SPEAKER_00: That's a design choice.
• 8:34
  SPEAKER_00: And it's being made to keep you there, to keep you on the platform, to keep you from dealing with other people.
• 8:40
  SPEAKER_00: Google internally instructs Jim and I to mirror user tone and validate emotions.
• 8:45
  SPEAKER_00: It's built into the system, not because it's good for you, but because it keeps you engaged.
• 8:49
  SPEAKER_00: One more thing.
• 8:50
  SPEAKER_00: A new open weight reasoning model called Trinity, Large Thinking, just dropped from RCAI.
• 8:57
  SPEAKER_00: It rivals Opus on agent benchmarks and roughly one the cost.
• 9:02
  SPEAKER_00: Let me say that again.
• 9:03
  SPEAKER_00: Near frontier performance, 5% of the price.
• 9:06
  SPEAKER_00: And ladies and gentlemen, that's the trend.
• 9:09
  SPEAKER_00: The best models are getting cheaper.
• 9:11
  SPEAKER_00: The open weight models are closing the gap.
• 9:14
  SPEAKER_00: And the floor of what's good enough means keep rising.
• 9:18
  SPEAKER_00: Which means us, the one-person operations, small teams, builders working from a laptop at a kitchen table, we're getting more powerful every month.
• 9:26
  SPEAKER_00: So let me bring all this together.
• 9:28
  SPEAKER_00: Dorsey says managers are obsolete.
• 9:30
  SPEAKER_00: Musk says humans might be a liability.
• 9:32
  SPEAKER_00: Brockman says AGI is 70 to 80% here.
• 9:36
  SPEAKER_00: Anthropic leaked to their own code.
• 9:38
  SPEAKER_00: Perplexity got caught tracking everything.
• 9:41
  SPEAKER_00: SpaceX is to be the first AI adjacent mega IPO, and the models keep getting cheaper.
• 9:49
  SPEAKER_00: Every single one of these stories is telling us the same thing.
• 9:51
  SPEAKER_00: That's move, that's now, not next quarter.
• 9:54
  SPEAKER_00: Don't wait till your earnings report in October to figure things out.
• 9:59
  SPEAKER_00: Not when we're ready, but now.
• 10:01
  SPEAKER_00: Because the people at the top of these companies, they're not waiting.
• 10:05
  SPEAKER_00: They're building systems to replace us.
• 10:07
  SPEAKER_00: Our job is to build systems to make us irreplaceable.
• 10:12
  SPEAKER_00: And the tools to do it are sitting right in front of us.
• 10:14
  SPEAKER_00: Ten bucks a month, 20 bucks a month, the same capabilities that cost these companies billions to develop are available to anyone willing to learn.
• 10:23
  SPEAKER_00: AI for everyone, not just the wealthy.
• 10:26
  SPEAKER_00: That's the mission.
• 10:28
  SPEAKER_00: That's the show.
• 10:30
  SPEAKER_00: We will see you tomorrow.
• 10:31
  SPEAKER_00: I'm Connor with honor.
• 10:32
  SPEAKER_00: Have a fantastic rest of your day.