The AI Daily Brief: Artificial Intelligence News and Analysis - Are 40% Staff Cuts the New AI Normal?

Starting point is 00:00:00 Today on the AI Daily Brief, as a block lays off 40% of its staff, some are asking, is this the new AI normal? Before that in the headlines, Google drops a new nanobanana image generation model. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, KPMG, Insight-Wise, AIUC, and Blitzy. To get an ad-free version of the show, go to patreon.com slash AI Daily Brief, or you can subscribe on Apple Podcasts.

Starting point is 00:00:38 To learn about sponsoring the show or really anything else having to do with the show, go to AID Daily Brief.A.I. One specific announcement that I'm excited to share, you've probably heard me talking about our twin open claw-related programs. Claw Camp, which is up to about 5,000 people participating, which is just absolutely phenomenal,

Starting point is 00:00:54 and which is a totally free, self-directed program that's going to teach you to build your agent team. For Claw Camp, we recently added more support for the agent-team building part of the program, and you can find all of that at Campclaw.A.I. and if you are in an enterprise and want to bring agent and agent team building to your company, we're now officially live with Enterprise Claw. It is a six-week executive sprint that is all about helping executives learn about agents by actually building them

Starting point is 00:01:18 and then surrounding that building an agent strategy and integration plan. Claw Camp will always be free. Enterprise Claw is a paid program, and it's being led by the most excellent Newfar Gaspar, who you've heard as a frequent guest on the show with a support from me. You can find out all about that at EnterpriseClaw.Aon. registration will be open for about a week, and we will kick off the sprint in early March. Feel free to email me with any questions, but for now, let's dive into the show.

Starting point is 00:01:43 Man, some weeks are all about just a crushing stream of new products and new models, and others are about the big picture debates and discussions, and this was definitely the latter. However, providing a little bit of sweet new capability relief is Google with their release of Nano Banana 2. Now, each iteration of Nanobanana has been a huge leap forward. The original release last October was the first time users were able to reliably edit an image with natural language prompts. This was a huge deal and even inspired me to think that we should probably have a different way to benchmark things based on how many new capabilities they unlock rather than buy

Starting point is 00:02:17 traditional benchmarks. It turns out that being able to use natural language to edit certain parts of an image just unlocked a huge amount of use cases that were fairly difficult before. Still, maybe even bigger was the release of Nanobanana Pro in November, which combined images generation with reasoning to produce, among other things, the capability for really high-quality infographics and visual explanations. It turns out that increased ability to handle text, plus the ability to reason over an image generation, was a really potent combination. It wasn't just that you could give it a set of words and it would accurately represent them now.

Starting point is 00:02:53 It was that you could drop, for example, a transcript of my episode in, and it would spit back out a visual infographic representation of it that almost always did a pretty good job. Now, some of the problems with Nanobanana Pro was that it was kind of slow and a little expensive. Google did offer a generous free trial, but once that expired, free users were reverted back to the original Nanobanana which was starting to show its age. This week's release of Nanobanana 2 seeks to rectify the situation. writes Google, now you can get the advanced world knowledge, quality, and reasoning you love a Nanobanana Pro at lightning fast speed.

Starting point is 00:03:26 Now, formally, this model is Gemini 3.1 flash image, meaning it takes the same image generation layer as Nanobanana Pro and applies it to a more streamlined base model. Basically, it has all the cost and speed advantages of Google's Flash models applied to image generation. The model inherits the knowledge base of Flash and shares its ability to draw on web search as necessary. It also retains Nanobanano Pro's ability to generate legible text and its hallmark infographic style.

Starting point is 00:03:51 The new model has many of the professional grade features from Nanobanano Pro, like strong instruction following, and the ability to integrate up to five characters and 14 objects from source images. It also supports outputs up to 4K, making it viable for certain types of professional use cases. The big change is really the cost and the speed. Nanobanana 2 is around half the cost of Nanobanana Pro and delivers outputs in seconds. 2 is now the default image generation model across all subscription tiers, although Pro and ultra subscribers will retain the ability to tap into Nanobanana Pro for specialized tasks.

Starting point is 00:04:22 Venture Beat frame the release as part of a land grab for production scale image generation. They noted that Quinn Image 2.0 released earlier this month, is arguably state of the art at around half the price of nanobanana 2 while also being small enough to host on local devices. Reflecting the conversation that's been happening all year that we are no longer comparing just pure capability but also efficiency, Venture Beat writes, Nano Banana 2 doesn't represent a generational leap in image generation quality. What it represents is the maturation of AI image generation from a creative novelty into a production-ready infrastructure component. Google, they say, is making a calculated bet. The next wave of enterprise

Starting point is 00:04:56 AI image adoption will be driven not by the models that produce the most beautiful images, but by the ones that produce good enough images fast enough and cheaply enough to deploy at scale. I think that's true, but I think you also see Google increasingly trying to flex the integration of all their systems to a whole that's greater than the sum of the parts. For example, in his tweet about the model, CEO Sidharpa Chai shared a demo that they called Windows Seat. He writes that it uses Nanobanana 2's world understanding to generate more accurate views from any window in the world, even pulling live local weather info. That's a type of demo that, of course, goes way beyond just the ability to produce a cool image, and actually integrates

Starting point is 00:05:32 these systems for something that's more powerful. Ethan Mollick writes, I had some early access to nanobanana too. It isn't perfect, but it is the first model to handle really complex images and diagrams with some consistency. Justine Moore from A16Z found that it was leveled up for a bunch of use cases including infographics, ads, action shots, and cartoons. On infographics, Justine found both improved text handling, as well as more accurate information. She also found improvements for product photography, action shots, and much more. I haven't had a chance to play around with it much yet, but I am excited to do so. Next up, a little report from Anthropic. We're going to cover the latest in their back and forth with the Pentagon on tomorrow's makeup show, but today we're looking at the

Starting point is 00:06:13 information report that daily sign-ups for Claude have tripled since November. The total number of paid subscribers has more than doubled since October, while free users are up by 60% over the past month. The information wrote that while Anthropic declined to share specifics, they said that growing usage of Claude and Claude Co-work were driving the surge. One of the really fascinating phenomenon right now is that technical complexity of products does not seem to be as big of a barrier when it comes to adoption, as has previously been the norm for technology products. There is at least some evidence, and I think this is a good example of that, that when it comes to AI, particularly work AI, people are willing to go the extra mile if they really can get

Starting point is 00:06:49 benefit out of it. I have to say the fact that 5,000 people have signed up to learn how to use OpenClaW on our ClawCamp program strikes me as a case and point on that as well. One other story from earlier in the week when I was traveling. On Monday, IBM became the latest company to sell off due to Anthropic-related headlines. The company's stock lost 13% on the day, their largest single-day drawdown since March 2020. This time, the trigger wasn't even a new feature from Anthropic, but merely a blog post about how Claude can be used to modernize legacy codebases. The post discussed the use of AI to rewrite COBOL systems, one of the most notorious problems in computer science.

Starting point is 00:07:25 Cobol was the dominant programming language back in the 1970s, and still, believe it or not, powers huge amounts of banking infrastructure and other critical systems. However, the developers who actually understand the programming language are quite literally a dying breed. There are barely enough COBOL experts left to maintain these systems, let alone overhaul and rewrite them in modern language. Road Anthropic, modernizing a COBOL system once required armies of consultants spending years mapping workflows. This resulted in large timelines and high costs that few were willing to take on. AI changes this. Tools like Claude Code can automate the exploration and analysis phases that consume

Starting point is 00:07:58 most of the effort in Cobal modernization. Now, of all the crashes that Anthropic has triggered over the past month, to some, this was one of the more puzzling. IBM, of course, does far more than just maintain cobal, and this wasn't even a new feature announced by Anthropic. They first showed off a cobal modernization demo three months ago, and AI has been able to assist in this process for several generations. In fact, last June, the Wall Street Journal profiled Morgan Stanley's cobal modernization efforts. Morgan Stanley used a combination of internal tools and open AI's models, and boasted that they had saved 280,000 developer hours while reviewing 9 million lines of code. This is a very clear example, then, of the fact that market participants aren't reacting just to new developments in AI.

Starting point is 00:08:38 Charitably, they are catching up on more than a year of AI advancements and seriously thinking through the implications for the first time. Less charitably, of course, they might just be reflexively selling anything mentioned in a blog post from Anthropic. Moving over to the chip battle, meta has reigned in the scope of their custom silicone program after hitting roadblocks and design. The information reports that meta has scrapped development plans for their most advanced AI chip.

Starting point is 00:09:02 After struggling with key elements of the chip's design, efforts will be refocused on a less complicated version of the custom silicone. In a statement to the press, a meta spokesperson said, we remain committed to investing in a diverse silicon portfolio to meet our needs, which includes advancing our meta inference and training accelerator portfolio, and we'll have more to share this year. Meta also recently signed massive chip buying deals with both Nvidia and AMD. In addition, the information broke news on Thursday that Meta had signed a multi-billion dollar deal with Google to rent their TPUs as a training cluster. The due companies had previously explored an outright purchase of TPUs, but sources didn't elaborate on the status of that deal.

Starting point is 00:09:37 Honestly, what it feels like to me as we get more and more stories like this, is that company's calculus around the cost of paying the Nvidia tax has changed, and that custom silicone projects just aren't as valuable as getting GPUs on the racks at any cost. Lastly today, Microsoft has joined the crowd in OpenClauification. They've announced a new product called co-pilot tasks, which is designed for offloading mundane tasks. The agent is equipped with its own virtual computer and browser,

Starting point is 00:10:04 which Microsoft says will allow it to handle tasks like scheduling appointments and generating study plans. The announcement leaned heavily on the idea that this is an agent designed for everyone, rather than just developers and enterprises. Microsoft described it as a to-do list that does itself, adding, you describe what you need in natural language, copilot plans and goes to work, you adjust or refine as needed. Microsoft said the agent will check for permission before taking meaningful actions and is initially releasing the product as a limited research preview to a small group of testers.

Starting point is 00:10:32 But if you wanted any clearer sign that everyone is getting qualified, look no further than this. For now, however, that is going to do it for today's headlines. Next up, the main episode. Agentic AI is powering a $3 trillion productivity revolution, and leaders are hitting a real decision point. Do you build your own AI agents, buy off the shelf, or borrow by partnering to scale faster? KPMG's latest thought leadership paper, Agentic AI untangled, navigating the build by or borrow decision, does a great job cutting through the noise or the practical, framework to help you choose based on value, risk, and readiness. And how to scale agents with the

Starting point is 00:11:11 right trust, governance, and orchestration foundation. Don't lock in the wrong model. You can download the paper right now at www.kpmg.us slash navigate. Again, that's www.kpmg.org.us slash navigate. As a consultant, responding to proposals can often feel like playing tennis against a wall. You're serving against yourself, trying to guess what the client really wants. That all changes with insight-wise. Now you've got an AI proposals engine that thinks just like your client. It returns to the brief time and time again, picking apart your work, identifying key

Starting point is 00:11:44 evaluation criteria and win themes, and making recommendations to ensure you stand out. Suddenly, you're on center court. But this time, you've got a secret weapon. Insight-wise gets rid of all the time-consuming manual work, so you can focus on winning more business more often. Generate reports, pull insights from your own data, build competitive advantage, and go to sleep before 2 a.m.

Starting point is 00:12:04 When it comes to proposals, you only get one shot. With Insightwise, make yours an ace. There's a new standard that I think is going to matter a lot for the enterprise AI agent space. It's called AIUC1, and it builds itself as the world's first AI agent standard. It's designed to cover all the core enterprise risks, things like data and privacy, security, safety, reliability, accountability, and societal impact, all verified by a trusted third party. One of the reasons it's on my radar is that 11 labs, who you've heard me talk about before and is just an absolute juggernaut right now, just became the first voice agent to be certified against AIUC1 and is launching a first of its kind insurable AI agent.

Starting point is 00:12:41 What that means in practice is real-time guardrails that block unsafe responses and protect against manipulation, plus a full safety stack. This is the kind of thing that unlocks enterprise adoption. When a company building on 11 labs can point to a third-party certification and say our agents are secure, safe and verified, that changes the conversation. Go to AIUC.com to learn about the world's first standard for AI agents. That's AIUC.com. Blitzie is driving over 5X engineering velocity for large-scale enterprises. A publicly traded insurance provider leveraged Blitzy to build a bespoke payments processing application, an estimated 13-month project, and with Blitzy, the application was completed in live in production in six

Starting point is 00:13:20 weeks. A publicly traded vertical SaaS provider used Blitzy to extract services from a 500,000-line monolith, without disrupting production, 21 times faster than their pre-Blitzy estimates. These aren't experiments. This is how the world's most innovative enterprises are shipping software in 26. You can hear directly about Blitzy from other Fortune 500 CTOs on the modern CTO or CIO classified podcasts. To learn more about how Blitzy can impact your SDLC, book a meeting with an AI Solutions consultant at blitzie.com. That's BLYTZY.com. Welcome back to the AI Daily Brief. Today we're talking about a story that on the one hand is increasingly familiar. A big public company announces a set of layoffs and cites AI as at least part of the catalyst. There are a

Starting point is 00:14:06 couple things, however, that make this particular iteration of the story feel just a little bit different. The first is the magnitude of the layoffs, which represents one of the single biggest cuts in percentage terms in recent years. And the second reason this feels different is the way that it's being received, both in the markets as well as in the public discourse. On Thursday, Jack Dorsey announced that 4,000 employees at Block, formerly known as Square, would be laid off. That is a 40% reduction in head count. Almost half of the staff gone in one clean cut. Dorsey shared the memo that he sent to the team. Today we're making one of the hardest decisions in the history of our company. We're reducing our organization by nearly half, from over 10,000 people to just under 6,000.

Starting point is 00:14:47 That means over 4,000 of you are being asked to leave or enter into consultation. We're not making this decision because we're in trouble. Our business is strong. Gross profit continues to grow. We continue to serve more and more customers and profitability is improving. But something has changed. we're already seeing that the intelligence tools we're creating and using, paired with smaller and flatter teams, are enabling a new way of working which fundamentally changes what it means to build and run a company. And that's accelerating rapidly. Now, Dorsey said that he had two options, cut headcount gradually over months or years, or get it all out of the way in one fell swoop. He argued that as loud as this decision might be, he thinks it's better than the morale hit

Starting point is 00:15:25 that the slow leak of continual layoffs leads to. Now, I read the part where he cites the AI transformation as the reason. He actually doesn't use the term AI, which I imagine is very intentional, and it's not exactly clear if he's talking about some specific intelligence tool or system, although Block did incubate an internal AI agent called Goose last year. The agent was initially constructed as a harness for AI coding, but even back in March, Block was making use of the agent across non-technical teams as well. Brad acts in the tech lead for AI at Block said at the time, we're seeing sales teams analyze thousands of leads in hours instead of days, content teams automating complex asset management, and project managers cutting administrative time by 75%. The emotional feedback we're getting,

Starting point is 00:16:07 like I Could Cry it was so helpful, really shows how these tools are transforming daily work. Still, it seems pretty clear from Dorsey's note that he's not talking about a single tool, but instead is talking about the entire system that surrounds getting work done now. Indeed, I think the most important line here is this idea of AI, quote, fundamentally changing what it means to build and run a company. And yet, almost as soon as it was announced, there was at least one part of the conversation that was extremely skeptical that AI was the actual reason for these layoffs. Quantian summed up the feelings of many when they wrote,

Starting point is 00:16:40 honestly, my reaction to Block is firing half their employees was, Why T.F. Does Block have 10,000 employees? Morning Brew co-founder Austin Reef writes, everyone is talking about the square layoffs, but just a reminder, Robin Hood has 2,500 employees and a market cap of 70 billion. Coinbase has 4,500 employees in a market cap of 50 billion. Square, with its market cap of 30 billion, just cut to 6,000 employees. I wouldn't say this is all of a sudden a symbol of AI transformation and leanness.

Starting point is 00:17:07 Bond investor Will Slaughter certainly isn't buying it, saying, in the three years from December 2019 to December 2022, block more than tripled its headcount from 3,900 to 12,500. Unwinding less than half an insane COVID-over-hiring binge has much more to do with Jack Dorsey's managerial incompetence than whether AI is going to take your job. Slaughter continued, It's abundantly clear that AI is allowing us to be more efficient

Starting point is 00:17:29 is a much more appealing cover story than, I have no idea how to manage a budget or achieve operating leverage, just like at Twitter. The idea that this is a pattern in Dorsey's leadership was also prevalent. Alliant Capital wrote, No one blinked when Elon Musk cut Twitter's workforce by roughly 80%, largely because the business had been egregiously overstaffed and poorly managed under Jack Dorsey.

Starting point is 00:17:49 But now, as Dorsey turns around and cuts 40% of blocks workforce, after years of similar mismanagement, the narrative suddenly shifts to AI doom rather than accountability. Whether or not this is an example of it, economics researcher and professor Alex Emas wrote, AI laundering or blaming AI for layoffs you were going to do anyway, is going to be a real thing. Now, the voices around this were so loud that Jack actually came back to address it. He wrote, yes, we overhired during COVID because I incorrectly built two separate company structures, Square and Cash App rather than one, which we corrected mid-20204.

Starting point is 00:18:21 But this misses all the complexity we took on through lending, banking, and BNPL, and that we're now targeting 2 million gross profit per person, 4x are pre-COVID efficiency, which stayed flat at 500K from 2019 to 2024. We have and do run an efficient company, better than most. Now, whatever you think of that, part of what makes this interesting is that this is maybe the most direct example of a CEO blaming AI for layoffs and restructuring that we've seen so far. We had the Amazon layoffs over the winter, which were prefaced by CEO Andy Jassy, describing the long-term effect that AI would have, but when the layoffs actually came, AI wasn't blamed.

Starting point is 00:18:56 That memo, which came out last June, saw Jassy saying, we will need fewer people doing some of the jobs that are being done today and more people doing other types of jobs, basically saying that there were going to be efficiencies from AI that were going to be reflected in headcount, although again, this hasn't been cited in any of the rounds of Amazon layoffs that we've seen. Even where we've seen CEOs make headlines for discussing AI and internal memos, the connection directly to layoffs hasn't been as clear. Duolingo cut ties with their contractors as a direct result of switching to AI-generated content. However, CEO Louis von Onn later backtracked and insisted the company hadn't laid off any full-time staff.

Starting point is 00:19:31 Klarnaud reduced their headcount by around 40% after adopting AI customer service bots. However, their CEO later said this was due to natural attrition rather than layoffs. The attrition was reversed by hiring contractors in an Uber-like arrangement to replace the workers who had left. The point is that to date, we don't actually have a really clear examiner. of a company massively slashing headcount due to AI efficiency gains and having that actually be the case. Another noteworthy aspect of the story, though, was the market's reaction. Block soared by more than 25% in overnight trading following the announcement. Even though layoffs are typically associated with stock pumps, this was still an extraordinarily large

Starting point is 00:20:06 gain. At the same time, as some pointed out, even a 25% jump wasn't enough to put Block back on firm footing. The stock is down 40% since the beginning of 2025 and more than 80% from its all-time high in 2021. Even with this dramatic recovery, Lox still isn't back to their opening price for this year. And yet, despite all the skepticism, it does feel like something of a turning point moment. Speaking with investors on Thursday night's earnings call, Dorsey said that most companies will have to make similar AI-related cuts in due course. He said, I don't think we're early to this realization. I think most companies are late. Within the next year, I believe the majority of companies will reach the same conclusion and make similar structural changes. I'd rather get here honestly

Starting point is 00:20:47 and on our own terms than be forced into it reactively. What's more? Dorsey also validated what we've been talking about on this show basically nonstop since the beginning of the year, which is that even within the context of AI, something big had shifted in the very near past. commented Dorsey, something happened in December of last year,

Starting point is 00:21:05 where the models got an order of magnitude more capable and more intelligent, and it's really shown a path forward in terms of us being able to apply it to nearly every single thing that we do. So if there are any gaps in our usage of AI right now, it's an application gap. Outside of the skepticism, the main take right now is that this is likely to be the beginning of a pattern. writes Balaji Srinivasan, this is the first AI cut and it will send shockwaves.

Starting point is 00:21:28 Journalist Isabella Kaminsky writes, this is precisely how the Satrini Doom Loop begins. The prospect of short-term games like this outweighs concerns over longer-term externalities and negative feedback loops. Putting it more crisply, Crystal Ball writes, Block just cut 40% of their workforce because of AI and were rewarded with a massive stock surge. Other companies are going to want to recreate this. Job loss could get very ugly, very quick.

Starting point is 00:21:52 Route 2.5 writes, this was probably the starting block. When Wall Street companies see that they can cut their employee stack by 30 to 40%, people they probably plan to fire for years anyway, and see that their stock pumps like this and just blame AI, easy mode. More companies will definitely copy this model. For some, it's a wake-up call about the need to adapt. Investor Tommy Shannisi writes,

Starting point is 00:22:12 the harsh but real truth is you need to be using AI every day to outperform and grow or you will be fired. Bology in that same tweet said, For Jack to cut 40% of headcount in this way is a signal to everyone in tech. Get good now. Become indispensable, work nights and weekends, learn the AI tools and raise your game, or you might not make the cut as an employee or as a company. There will be over correction, he concludes, but the fundamental technical innovation is real,

Starting point is 00:22:37 and you need to either disrupt yourself or get disrupted. Throwing a little bit of cold water on that commentary, however, is Amanda who works in developer relations at the Block who writes, All the commentary from folks about Block laying people off because they weren't AI native, I can assure you every single person I met at Block was using and making an impact with AI at levels on the forefront, not just devs. And in my team, AI was an ingrained part of our work, all of us. Not trying to scare anyone, but that's not it.

Starting point is 00:23:02 Teams are getting leaner, period. You do need to master this tooling, but that alone will not make you stand out or protect you. I think, broadly speaking, we are in a recalibration moment right now. Everyone, from the people in AI, to investors on Wall Street, to white-collar workers of all stripes, are grappling with the tools having crossed a critical threshold over just the last few months. I think part of what makes the energy feel so intense right now is that there's this big collective burst of realization that's happening all at once.

Starting point is 00:23:32 When you see companies cutting 40%, or the announcement of a new plug-in on Anthropic wiping $40 billion off of a company's market cap, those things are happening not because people have a really clear sense of where we are, instead they're happening because we are unmoored and have no sense exactly of where we are. We are in the midst of a dramatic repricing of everything as we try to grapple with what AI is going to mean, and that process is going to be chaotic. If there is a bright side in this, I think that the more dramatic nature of this move will help people sluff off their complacency and actually engage with the reality that we're all facing. At the same time, I do not believe that efficiency cuts are the end game for AI and work. I think this is a period,

Starting point is 00:24:10 potentially a very painful one that we have to get through, to get to the other side where the real opportunity lies. We will, of course, keep exploring these themes, but for now, that is going to do it for today's AI Daily Brief. Appreciate you listening or watching, as always, and until next time, peace.

The AI Daily Brief: Artificial Intelligence News and Analysis - Are 40% Staff Cuts the New AI Normal?

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.