A tremendous amount of news happened this week. Just to recap a few of the highlights:
OpenAI released Sora, their generative video social media app
Stripe announced instant checkout, integrated directly into ChatGPT
Jeff Bezos thinks we may be in an industrial AI bubble
And, Jeff Bezos thinks we’re going to build data centers in space
I’d highly encourage people to listen to a bit of Jeff Bezos’ viewpoints on AI at the current moment. His general take is that we’re in an industrial AI bubble similar to the telecom bubble of the late 90s/early 2000s. This bubble has the potential of popping, but similar to the telecom boom and bust, the infrastructure being built out today will more than likely be used for decades on end. And beyond that, Jeff Bezos is unequivocal about his excitement about AI.
He believes, “the biggest impact that AI is going to have is, it is going to affect every company in the world. It’s going to make their quality go up, it’s going to make their productivity go up. By every company I literally mean every company, it’s going to affect every manufacturer, every hotel, every consumer products company, etc. etc. etc.”
I talked last week a little about credible speakers, and I think that Jeff Bezos certainly would count as one. Having built out the dominant platform of the prior technology cycle, he certainly qualifies as an individual where will meets skill, and just given his position in life, he certainly operates in an information and access flow unavailable to 99.99% of people.
Which makes me think. What do Jeff Bezos, Jensen Huang, or Sam Altman for that matter know that you or I don’t? Why are they all generally promoting the same AI native/energized future? What are their incentives to promote this “augmented intelligence” future or to wax poetic about how every industry will be transformed? Will they make money if this future comes to light? Would they lose money if they’re wrong? Does it serve their egos to be put into the limelight, lauded, and treated like celebrities? Is this type of public proclamation par the course? Or an abberation?
Everyone should do their own research and come to their own conclusions, but it’s important to ask yourself why they’d be wrong about this specific type of future. It’s abnormal to predict that the way we structure society is going to be fundamentally transformed, but increasingly, that seems to be the consensus in the tech ecosystem spreading from the the major players at the table, to the surrounding participants writ large.
Parallel Intelligence
Simon Willison is a technologist I follow who has a widely read blog. He’s the co-creator of Django, a python web framework used by millions of developers. In sports parlance, he’s a bit of a beast but outside of the tech/vc ecosystem he’s not well known. In any event, a recent blog post of his talked about his experience of “embracing the parallel coding agent lifestyle.” Coding agents are the pacesetters of the AI application layer and they are widely cited as the perfect example of the shift of labor services into software systems, which is the broad theme I’m currently investing behind.
Simon’s blog mostly focuses around tactics on how to adapt to/optimize parallel AI workstreams, but it made me think - what will this look like in other job functions and when will it occur? Coding agents have exploded out of the gate for two reasons in my opinion 1) there are massive data sets of historical code that are readily available for an LLM to be trained against 2) new companies/products in the space (i.e Cursor, Claude Code, Codex, etc.) absolutely nailed the 3% rule of taking something that exists and is familiar (i.e the command line or a code editor), and adjusting it on the margins to create something completely transformative.
For most white collar jobs, there is no historical data set of clicks, so figuring that out will take time, but rethinking how people interact with email? Or the average website? Or insert the CRM that your industry uses? There’s a tremendous amount of room for those product surface areas to be rethought for the AI age, and it’s only a matter of time before these types of products become ubiquitous across the economy.
Now, to come back to my original point, if the future of work is parallel. What does that look like for your average white collar worker? And what will software look like in that environment? I’m going to overfit my view of the world for VC, but I think a lot of my perspective on this point is generally applicable:
1) I think that the future state of parallel work, will be by nature default collaborative. Most of office software today is single player by default. Yes there is collaborative tooling between internal and external parties, but most applications (i.e email, google docs, your CRM) start and focus with actions in single player mode. This will be flipped on it’s head in two ways 1) new software systems will need to be default collaborative by design 2) the default collaboration will need to focus on collaboration between humans and AI agents first, and human to human collaboration second. I could see this default collaboration coming to light in the form of a “News Feed” or “Chat” where my AI agents (when prompted) provide me with updates or suggestions on what I need to do. I could easily see an AI agent providing me with suggestions on how to make contact with a startup founder that I may have been chasing for 6 months to no avail (happens all the time), and potentially providing me with research on that particular founder/company that they’ve created over night.
2) Output Review is going to be critical. The whole premise of AI is that you’re doing more with less, and that you’re more efficient with your time. The example with software engineers is they’re no longer taking 16 hours to generate and review 1000 lines of code, they’re skipping to the end and spending 4 hours to just review 1000 lines of code. The same thing will be true with most white collar work. Whether you produce memos, decks, or just live and die within email - the future of white collar work will be out output review rather than creation, and as a result, the ideal parallel software system will make output review incredibly easily and intuitive. For me this is where that 3% idea comes in, I think powerpoint, word, and excel are all pretty well optimized. I don’t think I need, or even want, a better version of these tools. The innovation will actually just entirely come from finding a way to natively bring these experiences into a single application that also has access to all of my necessary context + functionality that allows for making updates on the fly incredibly easy.
3) Everything will fall apart without priority ranking.One thing I absolutely know to be true about the world is that priorities shift. This applies to work and to life. At 25, my priorities on a friday night were to go out on the Lower East Side or West Village and see what shenanigans I could get into. At 30, my priorities are to maaaaybe watch one more episode of Black Rabbit (fantstic show btw, would recommend). So if priorities shifting are just fundamentally a fact of life, you’ll need to find a way to communicate that to your team of AI agents. This is why I think priority ranking, is going to be a huge deal. AI agents are, as far as I know, responsive systems. They don’t proactively create and complete work, so having some sort of communication method where they know what to prioritize and for how long is going to be important. From a product experience perspective, I honestly just view this as a simple dashboard. I should have a single place where I can include my top 5 priorities, and I (in concert with my AI agents) will work on those in parallel till they’re completed.
If I take a step back and think about the features I’ve described, I think about it from a design perspective as email meets powerpoint meets asana dashboard, with a unified data/context layer underneath. This vision for parallel work overfits for VC a little bit, but it should be broadly applicable to any other investment or sales role (workflows are generally the same, but different outputs, complexity and industry jargon).
Btw Black Rabbit is a pretty good show. Is it the best TV you’ll ever watch? No, but it is an entertaining limited series with Jason Bateman and Jude Law. NYC is the backdrop of the series. I’d give it a 7 out of 10. See you all next week, and don’t be afraid to like, subscribe, or share All Things Venture. Peace!
Item 3 resonates with me. My short and long term goals drive how I prioritize my actions each week, and I work across multiple platforms (e.g., google docs, Microsoft ppt, Gmail/outlook, obsidian, etc.) that capture my progress. Integrating the data across all these tools and coupling it with AI agents/inferencing would tremendously benefit my productivity and level of insight into how I'm tracking against my goals. Hope you, Tess, and the family are doing well. Have a great week