What does Dan predict about Codex and Claude Code?
They'll act as the primary work environment—an OS where humans and agents collaborate, run SaaS tools, and share context across tasks.
Video Summary
Codex and Claude Code (cloud coding environments) will become primary work platforms where humans and agents collaborate.
Most companies will run a shared 'super-agent' (likely in Slack) that employees use daily rather than many personal agents.
SaaS is not dying—agents will increase SaaS usage and open new margin models where users bring AI tokens.
PMs, full‑stack designers, and forward‑deployed engineers are the roles most likely to thrive in the AI era.
Automation doesn’t eliminate human work—agents need human oversight; staying relevant means experimenting and 'riding the models.'
They'll act as the primary work environment—an OS where humans and agents collaborate, run SaaS tools, and share context across tasks.
No—Dan argues agents will increase SaaS usage, creating high-volume agent-driven interactions and new economics like user-supplied AI tokens.
Product managers, full‑stack designers, and 'forward‑deployed' engineers—people who combine judgment with hands‑on use of models and agents.
Dan says the CLI era was short-lived for broad workflows—graphical agent-centric environments are taking over for most day-to-day work.
Experiment with models and agents ('ride the models'), build practical workflows around them, and focus on creative, higher‑value work that leverages AI.
Not necessarily—automation often increases output and requires human oversight; agents create more review, integration, and quality-control work.
"The AI job apocalypse is not really a thing. I am super bullish on PMs and full-stack designers."
"Automation is a lie. Every agent needs a human."
"It's going to bifurcate in two main ways."
"I think the SaaS apocalypse is dumb. I would buy SaaS stocks right now."
"We speed ran the CLI era. It was nice while it lasted, but I think CLIs are over."
"WorkOS turns deal blockers into drop-in APIs with a modern developer platform built specifically for B2B SaaS."
WorkOS provides essential infrastructure for startups looking to integrate features demanded by large enterprises, such as single sign-on and audit logs.
It offers APIs that streamline the integration process, making it easier for companies to build products that meet enterprise needs.
The platform is rapidly becoming an essential tool for any startup aiming to expand into the enterprise market, likened to "Stripe for enterprise features".
Startups that utilize WorkOS can achieve enterprise readiness more quickly, which allows them to focus on growth and innovation.
"One of my favorite questions because I think if you look at the benchmarks, AI is going to take all of our jobs."
The conversation outlines three predictive buckets regarding the future of work: changes in work processes, the changing nature of tasks, and the profiles of successful individuals in this new work environment.
There is a consensus that while AI's capabilities are advancing rapidly, humans will still have significant amounts of work to accomplish, despite technological progress.
Critical predictions include the idea that work will split into two main streams; one where employees utilize AI agents to delegate tasks, and another where their workflow will integrate with AI-supported environments.
"The way you will be doing work in a year is going to bifurcate in two main ways: how you use agents and your workflow environment."
The future will see employees engaging with AI agents as part of their daily routines, creating a division between human tasks and those delegated to AI systems.
Most individuals will likely interact with at least one dedicated agent, possibly through platforms like Slack, to offload various responsibilities.
The work environment will evolve around integrated systems like Codex or Cloud Co-work, positioning them as primary operational frameworks for various tasks.
"There's this interesting shift towards having one super agent for the entire company rather than personal agents for everyone."
Initial excitement around personal AI agents has shifted towards a model where companies may favor a centralized "super agent" that manages broader tasks across the organization.
This transition comes from recognizing the maintenance burden personal agents impose on users, prompting a need for an operational structure with one dedicated agent that handles shared tasks.
The prediction includes a future where individual agents could emerge as AI technologies improve, allowing for more personalized assistance without the heavy operational demands currently involved.
"Agents need people who care about them; if you sever that connection, they become less useful."
The utility of AI agents is closely linked to human engagement; they require oversight and interaction to remain effective and relevant within an organization.
As organizations evolve, the success of particular AI agents may depend not only on their technological sophistication but also on the human elements that guide and manage them.
The ongoing relationship between users and AI agents will determine how effectively these tools can be integrated into daily workflows, with a focus on collaborative improvement over time.
"These use cases get more specialized, and agents become less fiddly. They just work better."
The development of AI agents is progressing towards greater specialization, resulting in improved functionality and user experience.
A key application area for these agents is in workplace communication tools, particularly Slack, where they are expected to enhance productivity.
Users generally prefer to keep their personal and work tasks separate, indicating a potential shift in how AI agents are utilized in different domains—work versus personal life.
"There's this whole territory of using personal agents for your computer errands, like ordering groceries."
The concept of personal AI agents managing everyday tasks, referred to as "computer errands," is gaining traction.
These tasks might include various household chores such as grocery shopping, highlighting the broader applicability of AI beyond just workplace environments.
Despite the broader focus on personal tasks, the primary attention of AI development seems to remain on improving work-related functionalities.
"They created this super-powerful coding paradigm."
The introduction of AI coding tools, such as Claude Code and its development into Co-work, represents a significant advance in how coding tasks are performed.
Originally, these AI agents ran locally on computers, granting them access to full system capabilities and enabling a more effective use of programming commands.
As these AI coding agents improve, they not only serve as coding assistants but also facilitate a wider range of work-related tasks, making them versatile across various applications.
"OpenAI has surpassed them recently, but they were very early on this."
The competitive landscape between AI models, particularly between OpenAI and Anthropic, reflects a race in the development of advanced coding capabilities.
Anthropic's early focus on local coding agents provided them with a head start before being surpassed by OpenAI's iterative advancements.
This rivalry continues to drive innovation, with each company working to refine their coding agents to better meet user demands for efficiency and usability.
"Codex has a browser in the app that lets it see everything you're doing."
The integration of a browser within the Codex app allows for seamless interaction between the user and the AI, creating a more cohesive work environment.
Users can manage documents, conduct research, and have real-time assistance from the AI, enhancing productivity and reducing the time spent on tasks.
The combination of AI functionality and user interface design is essential in making these tools beneficial and easy to use in practical scenarios.
"The SaaS tools will run within Codex or Claude Code."
A shift is occurring wherein software as a service (SaaS) tools are beginning to operate within AI environments, changing the traditional approach to their integration.
This new paradigm involves users bringing their AI to these platforms rather than the other way around, highlighting the user's control over both the tools and their data usage.
Developing SaaS applications that are friendly to AI agents opens new avenues for efficiency, allowing for a dual-use design where both human users and AI agents can work simultaneously on the same platform.
"Cursor has more distinctly chosen a lane, prioritizing being a tool for programmers."
Cursor distinguishes itself by specifically targeting programmers rather than aiming for a more generalized user base. This focused approach may limit its potential reach in some areas but still recognizes the expanding definition of a programmer, opening up a significant market opportunity.
The evolution of AI platforms emphasizes the necessity of robust integration with cloud computing. As companies develop sophisticated models, they realize that optimizing interaction with these models requires a well-defined harness or framework.
The shift is evident in how platforms like Anthropic and OpenAI are moving towards advanced cloud-managed agents, which facilitate better overall performance from AI models compared to merely using prompt-response systems.
"We're moving into a new paradigm where the human and the agent are working together on the same piece of work."
The transformation involves a new collaborative dynamic where human users and agents operate concurrently on tasks rather than sequentially. This collaboration necessitates innovative software designed to provide visibility into both the agent's actions and the user's inputs.
Traditional productivity software primarily designed for human use is evolving to incorporate agents that can perform multiple functions simultaneously, drastically simplifying product design and user experience.
As agents take on more responsibilities, products will need to adapt to facilitate quick approval processes, effective logging, and rollback capabilities to handle the faster-paced interactions driven by AI.
"CLIs are over. We speed ran the CLI era; it was nice while it lasted."
Though command line interfaces (CLIs) have a long-standing presence in tech, the rise of user-friendly graphical user interfaces (GUIs) tailored to work with AI agents is reshaping the interaction landscape.
The conclusion is that while CLIs will persist, many technical users are gravitating towards utilizing integrated tools like Codex and Cloud Code, suggesting a significant shift away from CLI dependency for daily tasks.
This new environment supports applications that operate seamlessly within these advanced tools, fostering a faster and more efficient workflow for both developers and everyday users.
"When Codex interacts with another agent, it can provide context about me that I would not be able to type."
The interaction between multiple agents can significantly enhance productivity, as they can communicate more efficiently than a human user could when conveying context or conveying detailed instructions.
This outcome highlights the potential for faster problem-solving and increased effectiveness in products that allow for integrated agent collaboration. Real-time synergy between agents can streamline tasks and feedback loops in ways that traditional methods could not achieve.
The movement towards multi-agent systems underscores the anticipated evolution of user interfaces, making AI assistance a more integral and seamless aspect of day-to-day workflows across various platforms.
"The whole paradigm starts to change when you assume that everyone's got an agent, and those agents are talking to other agents in a really magical and important way."
When developing agent products or new software experiences, the process changes fundamentally if you assume every user has an agent such as Codex or Co-work. Instead of a lengthy onboarding checklist, users can simply input a prompt which allows their agent to communicate directly with the app.
Codex can leverage vast amounts of user data to create personalized experiences, sharing relevant information and adapting to individual user needs effectively.
The ability to instruct Codex to diagnose and rectify issues when they occur represents a significant shift in how technical products operate.
"SaaS companies are going to see an insane spike in the amount of demand they have because there’s going to be tons of agents using these products at a very high volume."
Emerging trends suggest that Software as a Service (SaaS) companies will benefit from the integration of agents within their models, leading to enhanced user engagement rather than reducing it.
With AI tools acting as collaborative partners in using SaaS applications, demand for such services is expected to surge, thus potentially preserving company margins amidst rising operational costs associated with token usage.
The conversation reflects a bullish sentiment on the future of SaaS, predicting that companies will maintain their significance despite the rapid advancements in AI.
"Every time you automate something, in order to ensure that the automation is working well, you need a human on top of it."
The discussion highlights a critical understanding that, even in an increasingly automated environment, human oversight remains indispensable. This suggests that many positions, particularly managerial roles, are likely to continue to evolve rather than disappear.
Managers engage actively with their teams, reflecting the human element that is essential for effective productivity. Automation may simplify tasks, but the complexity of ensuring effective management increases.
Notably, dependency on benchmarking may mask the true autonomy level of AI tools, indicating that while much automation is in place, the workload on human employees may not significantly reduce.
"When we get to that point, I will be very easy for me to change the benchmark to zero out the current model."
A personal benchmark experience was shared, detailing the comparative performance of AI against human engineers. While early models struggled, indications are that advancements in AI, particularly with iterations like GPT 5.5, show notable improvement in problem-solving capabilities.
This iterative process of benchmarking AI performance reflects ongoing efforts to calibrate AI’s capabilities and expose areas of potential growth.
Future iterations of AI models are anticipated to close the gap significantly with human engineers, as the architecture of models evolves to handle complex tasks more autonomously and effectively.
"Every coding model on the market will take that instruction seriously. If I tell it, 'Here's a bunch of issues, go fix it,' they will try to fix the issues."
The speaker reflects on their experience directing an AI coding model to resolve multiple reported issues within a production environment.
They believe that AI models will take such instructions seriously, attempting to address indicated problems effectively.
In contrast, a human senior engineer conducts a more thorough analysis of the codebase, determining whether significant rewrites are necessary and acknowledging the inherent challenges.
The nuanced capabilities of AI models highlight their limitations compared to human engineers, especially when it involves understanding the broader context and implementing complex solutions.
"Benchmarks rise on problems that we've framed, that we can articulate, that we can score."
The discussion emphasizes that while AI models are becoming progressively better at automation, this does not eliminate the need for skilled engineers.
Benchmarks are often based on measurable, articulated problems, which can overlook valuable human work that defies easy quantification.
Even as AI continues to improve and take on more tasks, human intuition, creativity, and higher-level problem-solving remain irreplaceable in technology.
"The number of pull requests you get skyrockets, and a higher percentage of your company or your users are doing things that previously only technical users could do."
The evolution of work dynamics driven by AI has resulted in non-technical employees increasingly participating in tasks that were once confined to technical roles.
The influx of pull requests from a diverse range of personnel highlights a significant shift in collaboration and contribution within companies.
This innovation, while positive in expanding participation, introduces challenges related to managing and integrating new code effectively into existing systems.
"Now that everyone can do everything, there's just confusion about what the hell is my job anymore."
The transformation brought by AI is leading to a blurred understanding of individual job responsibilities, as professionals from various disciplines, like engineers, PMs, and marketers, find themselves engaging in overlapping tasks.
This trend creates uncertainty regarding the traditional definitions of roles, making people question their specific responsibilities in the workplace.
Over time, it is expected that this confusion will settle, and roles will normalize, with professionals continuing to focus on their specific fields while also integrating AI tools into their work processes.
"Automation was supposed to take away jobs, but it looks like it just created one or many."
The conversation highlights the rising concept of "forward deployed engineers," a role that bridges the gap between AI systems and human management.
Even as AI models become more powerful, they still require human oversight to ensure they operate correctly and efficiently, indicating a growing demand for professionals who are adept at managing AI systems.
This new role emphasizes the importance of human involvement in AI, where responsibilities include not only the initial setup of AI systems but also ongoing maintenance and optimization.
"More work means reviewing all this sloppy output."
With the proliferation of AI and automation, the volume of output increases, resulting in a greater necessity for data engineers who can sift through and ensure the accuracy of data-related tasks.
A significant challenge arises as data scientists find their roles shifting from analysis and insight generation to reviewing the work of others, necessitating robust systems and agents to filter and manage basic data queries efficiently.
Organizations that implement these supportive systems, such as data science bots, significantly improve workflow by allowing specialized teams to focus on complex analysis rather than mundane tasks.
"Your company's only going to go as far as your CEO goes in AI."
Despite the rapid changes in various tech roles, some positions remain largely unchanged, such as those of CEOs and middle management, who may not feel the immediate pressures of AI integration.
There exists a possibility for roles in sales to be less affected due to their inherently personal nature, focusing directly on engaging with potential clients.
The discussion points to a future where the integration of AI into all levels of leadership becomes essential for organizational advancement and efficiency.
"It's going to be a lot more reviewing of other people's output as a part of the work."
The work in sales and customer support is evolving with advancements in AI, particularly in how tasks are performed.
Sales processes have benefitted from AI, allowing for improved sourcing and qualification of leads, while the essence of a salesperson's role remains largely unchanged.
In customer support, the integration of AI has also led to fundamental changes, indicating a positive shift for these roles.
The future work environment will likely require employees to spend more time reviewing outputs generated by AI systems and managing these AI agents effectively.
"It's just an extremely interesting engineering challenge of building a system to enable everybody else in the organization."
As AI becomes more integrated into various work processes, there arises a significant engineering challenge—designing systems that allow less knowledgeable users to utilize AI effectively without errors.
This involves creating frameworks that can enable all team members to engage with AI tools responsibly, suggesting that the focus should be on creating user-friendly systems rather than merely monitoring AI outputs.
Professionals in technical roles can dive deeper into complex questions concerning AI, while non-technical users will interact with a system designed to accommodate their skills.
"We will be reading way more AI-generated writing in documents and emails, and we will like it."
People are likely to engage more with AI-generated writing in various contexts, including documents and communication.
The speaker shares a positive experience utilizing AI agents in planning and strategizing, highlighting efficient workflow and quality outcomes through AI assistance.
There is an acknowledgment that while AI-generated documents vary in quality, the expectation is for these to be high-quality and coherent. AI can produce content that meets or exceeds the capabilities of many human writers, which should alleviate fears around AI writing.
"The aversion to AI-generated stuff will go away because the kind of strategy document that GPT-5.5 can write is way better."
The perception of AI-written content is changing, especially in professional settings where utilitarian functions like planning and emails benefit significantly from AI assistance.
As AI writing capabilities improve, the fear associated with AI-generated content is expected to diminish as its practicality becomes evident.
The speaker emphasizes that human writers still play a crucial role, but AI can augment writing tasks effectively, especially when it comes to generating documents that require a combination of clarity and comprehensive detail.
"I am super bullish on PMs."
The speaker expresses strong confidence in the future prospects for Product Managers (PMs) as AI continues to develop and integrate into various sectors.
PMs who can harness AI tools will likely excel, as the evolving landscape favors those who are adaptable and can pair technical knowledge with strong product instincts.
The success of a PM named Marcus, who leverages AI tools effectively despite a modest technical background, exemplifies how critical it is for PMs to adapt and learn new skills in response to technological advancements.
"He can just do it. And it's super impressive, and it makes me very bullish on any PM who gets really AI-pilled."
The conversation highlights the shift in responsibilities for product managers (PMs), emphasizing their newfound empowerment through AI tools. These advancements enable them to focus on critical tasks such as identifying what to build and which problems need solving, rather than being bogged down by extensive organizational duties.
The sentiment is echoed with enthusiasm, as the speaker notes that this empowerment leads to a more dynamic, innovative environment where ideas flourish.
"If you're a designer, and you're in these tools all the time, you're so used to... now they can actually build it."
The role of designers is evolving due to the accessibility of new AI tools. Designers, now acting as full-stack professionals, feel empowered to execute their visions without relying heavily on engineers.
The freedom to create not only enhances the design process but also fosters a sense of ownership, where designers can produce and submit their work through pull requests, streamlining collaboration and reducing handoff delays.
"The AI job apocalypse is not really a thing."
The speaker challenges the narrative surrounding a potential job apocalypse due to AI. They assert that while companies are reorganizing, attributing this solely to AI overlooks other factors like overhiring.
The conversation suggests a stable demand for human roles, noting that new AI models contribute to commoditization, making traditional competencies less valuable. However, the need for creative humans who can leverage AI for unique applications remains vital.
"The only thing you need to do is ride the models."
To stay relevant in the job market, individuals must embrace and experiment with new AI models. This proactive approach allows workers to enhance their skill sets and integrate new technologies into their workflows.
The advice emphasizes the importance of curiosity and experimentation, suggesting that by continually exploring these advancements, workers can identify new applications for them in their specific roles and stay ahead in a rapidly changing environment.
"I think the edge of AI is wherever AI meets a real human doing something."
The conversation emphasizes that the forefront of AI innovation is not necessarily in tech hubs like San Francisco but rather at the intersection where AI technology interacts with real-world human applications.
Individuals outside these tech centers are often the first to uncover valuable uses for new AI models, as they apply these tools in their everyday tasks.
The speaker reflects on their experience in Brooklyn, suggesting that their proactive use of AI tools places them ahead of others in San Francisco who may be involved in AI development but lack practical application insights.
"No matter how much money you have, you have access to the most advanced AI model."
The speaker highlights the democratization of AI technology, noting that even individuals with limited resources can access cutting-edge AI models, barring a few exceptions for those directly employed in leading AI companies like OpenAI or Anthropic.
This accessibility contrasts starkly with historical technology trends where innovations might have been costly and restricted to large corporations, allowing broader participation and creativity in leveraging AI for various applications.
The idea that AI was developed within a culture aiming to make it "too cheap to meter" underscores the importance of its availability to everyone, fostering the growth of both small and large companies.
"On the one hand, a lot has not changed; on the other hand, every role has transformed."
The discussion touches on the paradox of evolving work environments due to AI, where traditional communication methods and tools, such as emailing and using Slack, remain in place despite the transformation of specific job roles.
Job functions have evolved, with different disciplines integrating AI differently; for instance, engineers no longer write code as they did, and product managers have shifted their approach to documentation.
The sense of continuity in certain workflows suggests that while AI alters individual roles, fundamental structures in professional life are resilient and adaptable.
"Some people believe that something incredible will happen that will change everything; others fear something terrible over the horizon."
The speaker reflects on the tendency for varying public perceptions regarding the future of AI, drawing a parallel to medieval interpretations of the unknown.
This perspective illustrates that people oscillate between optimistic and pessimistic views of advanced technologies, fearing dystopian outcomes while also hoping for utopian benefits.
As AI continues to evolve, the realization that each technological advancement introduces both positive and negative aspects is crucial to understanding the complexities of its impact.
"The best way to figure out interesting, useful things to do with AI is to do something enjoyable."
The speaker advises listeners to actively experiment with AI tools, suggesting they engage with their workflows using models like Codex or explore agent products to identify personal applications.
Finding joy in utilizing AI can enhance creativity and innovation, steering away from a fear-driven approach to technology use.
Encouraging a problem-solving mindset, he recommends that individuals look for challenges in their lives that AI could help address, ultimately leading to satisfying discoveries and solutions.
"AI is actually very good for quantum physics if you get into it."
The discussion ventures into the intersection of artificial intelligence and quantum physics. The speaker references a book titled "The Rigor of Angels," highlighting its exploration of significant historical thinkers such as Heisenberg, who is known for his uncertainty principle, and the Argentinian fiction writer Borges, whose stories are increasingly relevant to AI discussions.
The speaker finds the overlaps between these ideas and AI to be mind-blowing and highly recommends the book to those interested in these themes.
"My current obsession is The Power Broker."
The speaker shares their current reading obsession, "The Power Broker," which provides a compelling history of New York. They find it engaging and reflect on how it never seems to end.
In terms of entertainment, the speaker mentions enjoying basketball, particularly as a new Knicks fan, and recommends a documentary mini-series titled "The Dark Wizard," about extreme athlete Dean Potter. The documentary explores his adventurous psyche and parallels it with the mindset of founders and those pursuing high-risk endeavors.
"Codex... It's just really good."
The speaker describes their excitement about a specific AI tool, Codex, characterizing it as transformative for their work process. They emphasize that Codex is superior in comparison to other tools available, particularly for tasks like managing emails and checking analytics.
There is a strong endorsement for Codex, which the speaker believes would do a disservice if overlooked. They express a willingness to switch tools if another one becomes more effective in the future, maintaining an open-minded approach to using different AI tools based on their capabilities.
"When you're dealing with hard things, what you want to do is be able to relate to it from a position of spaciousness and strength."
The speaker shares a personal motto that they developed in college: "Do things worth writing about and write things worth reading." This reflects a purposeful approach to their work and life.
Additionally, they highlight the importance of maintaining a perspective of strength and spaciousness when confronting difficult challenges, particularly regarding fears about AI's impact on jobs. They encourage self-reflection to ensure a productive mindset when addressing such fears.
"The most useful thing you can do is find ways to use it well in your life and share it."
The speaker stresses the importance of engaging with AI positively and creatively, rather than getting caught in debates about its implications. They suggest that the joy and utility of AI can be maximized when individuals experiment and share their findings.
They believe that everyone should have fun with AI, as it enhances experiences when people collaboratively explore its possibilities.