Two crazy AI weeks
Reflection time
The last 2 weeks were wild. Both Anthropic and OpenAI are releasing a new wave of coding agent models, and they are wildly capable. Codex 5.3 in particular is my daily driver right now, and I have never seen things move so quickly.
Instead of boring you with the reactionary posts about the models like a typical tech influencer, I will be sharing some thoughts and observations on the direct and secondary effects on the field.
Younger Teams
Despite getting more capable models, all the AI players are hiring like crazy. The The Codex team in particular is shouting on Twitter about how they need more people to expand their team. Anthropic is hiring various different Go-to-market positions to transition into different verticals. These are great signs that our jobs are still here to stay. But the nature of the job is no longer about the design patterns or the algorithms. It’s much higher level now: a complete E2E architecture vision to wield the coding agents and realize such a vision.
But it’s not just the AI labs who are hiring; the consumer end is hiring as well. The interesting part about this portion of the job market is that enterprises realize younger, fresh grad candidates adopt AI much better and faster than their grumpy decades-old sysadmins. Newer Agentic Experience teams are being spun up with 1 experienced engineer leading groups of 0-2 years of experience with multiple coding agents running either on their laptop or on a cloud. Newer “cloud prompting” platforms are being spun up from these teams to create more secure, easier-to-use AI co-workers.
And the application is not limited to just coding!! It’s also office works like Business Intelligence, Marketing, Logistics, etc.
Deep Blue, Spark Orange
We coined a new term on the Oxide and Friends podcast last month (primary credit to Adam Leventhal) covering the sense of psychological ennui leading into existential dread that many software developers are feeling thanks to the encroachment of generative AI into their field of work.
We’re calling it Deep Blue.
On Lobsters, I advocated for an opposite term: “Spark Orange”.
It’s the joy of getting shits done.
No more the long-dreaded backlog of toils and ideas that you never have the time to build.
No more the back-and-forth design discussions that take hundreds of engineering hours.
Now, the moment you have a spark of an idea tinkled in your head, the coding agent will help you realize it immediately.
Ship your ideas! Now!
It’s never been a better time to build things.
Good Reads
There are a few interesting contents in the last few weeks:
Russ Cox commented on how Go project should handle AI-generated PRs.
The “Legal Concerns” section in particular is a fascinated read.Dwarkesh Podcast:
Interview with Elon on the necessity of Datacenter in Space.
My take from this is that Solar Energy is the most viable path toward sustainable (and green) energy in the next decade.Interview with Dario on the economic of LLM.
It’s very interesting to hear how Dario is so focusing on drug discovery and bio-risk. I think they are among many level of applications and risks that AIs have opened the door to, not entirely sure why he is worried about this specific set of ideas.
Latent Space interviewed Jeff Dean. Much more technical than the 2 interviews above. I need to watch this again a few more times, but I don’t think there is anything too surprising. I think it’s very clear that unlike other labs, Google are much more cost-conscious because they serve these models at a much bigger scale as they are rolling out the LLMs to more verticals. Those constraints spawned lower level innovations and Google is an absolute powerhouse at that.
OpenAI - Harness engineering: leveraging Codex in an agent-first world.
The blog opened with “Over the past five months, our team has been running an experiment: building and shipping an internal beta of a software product with 0 lines of manually-written code.” which is a wild statement. I personally suspect this is their Codex-tangent team cooking a GitHub replacement, but I have no insider info on this.Cursor - Towards self-driving codebases.
Tangent to this, the small paragraph about “Synchronization overhead” is what I have been thinking about at the back of my head the last few weeks. Let’s say I have 100 agents running in parallel and one of them decided to refactor//common/lib/foo → //common/lib/bar, then how do the other 99 agents know about this refactoring to adjust accordingly? When do they do a pull/rebase? What if there are conflicts? etc…Using go fix to modernize Go code.
Go team essentially created a tool to help the AI labs modernize their training data corpus. This means that the newer models can be trained with the latest Go versions syntax and best practices. This is probably the best way for a programming language to stay relevant in the age of LLM-powered coding agents.Peter Steinberger on joining OpenAI and building in EU
Grok translation:
In the USA, most people are enthusiastic.
In Europe, I get insulted, people scream REGULATION and RESPONSIBILITY.
And if I really build a company here, then I get to struggle with things like investment protection laws, employee co-determination, and paralyzing labor regulations.
At OAI, most people work 6-7 days a week and get paid accordingly.
With us, that’s illegal.
Also, mindset. How I found purpose again after 3 years of searching:
USA: "oh man this is so great let's build sth cool!"
AT: "Yeah, but take care of yourself, so you don’t get another burnout, okay? So slow down a bit."
As somebody who is living in the EU and is planning to stay here for the foreseeable future, I will be reflecting on this a lot. At the same time, I got to meet some incredible folks who are relocating to Amsterdam just this week! So I hope things will continue to get better from here on out.
Slow Down
With the current AI waves, I have never felt as much pressure or dread.
Yet come with them are a lot of joy and energy to keep on building new things.
It’s Lunar New Year, Tet for us Vietnamese, the time when we gather as a family and celebrate the past and look forward to the future.
It’s also my dear daughter’s birthday this week. She is turning two, and I am incredibly proud of her, and of my wife, our family, for what we have accomplished in the last 3 years to get where we are today.
I am taking this week off to reflect, to breathe, and to get some quality family time.
And I think you should too.
Pace yourself, for it’s a marathon, not a sprint.