Claude Code Is the Inflection Point
newsletter.semianalysis.com
newsletter.semianalysis.com
And to think we are limited by Compute Power even when we throw money at it, along with so many improvements in the pipeline and low hanging fruit. Scary to think what the future will be like, especially in terms of jobs. I have already seen some of our department cutting down 10-15% simply because something like dashboard now requires far less time to do.
HN often focuses on whether LLM cane take over or replaces the job of programmers, it is actually outside tech world which is getting hit the hardest.
Could anyone translate for someone who has never used an iPhone or followed particular versions of iPhone?
iPhone 5 ( and its Android counterpart era ) for many was the first Smartphone they use, we are talking about normal consumers, not the audience of HN. Even many of the so called professional YouTubers and Reviewers only have their first Smartphone during iPhone 4 - 5 era. Four years in many who thought it was hype, many who didn't understand why internet on your phone was a big thing, many who didn't know why you want an App. It also the first time iPhone was made with Supply Chain and operation in mind because previously supply could not keep up with demand. I believe it was first to break 100M yearly unit sold. iPhone 4 might have been luxury, but iPhone 5 was the time when majority wakes up and finally say we need smartphone. To put this in AI perspective, Claude Code is working, ( aka the PR commit percentage in the article ).
iPhone 6 Plus was the first iPhone with big screen at 5.5" ( 16:9 ), breaking 200M yearly sales, and finally a data points to the world is moving to big screen called phablet at the time. Smartphone is replacing Personal Computer, or your Smartphone is your PC replacing Desktop and Laptop. The Internet is happening on Phone, not PC. At this point it is just consistent growth, 200M unit sales as standard until around 2020 when things start to level off in 2020.
It is good for handling all code changes where there are no security aspects involved.
This job is clearly going to be modified in some way.
We finally have a legitimate way to query data using natural language and create dashboards. I don’t know what the point of learning sql is now - at least for olap users.
* we don't need to know how to write the code, because AI can write code.
* we don't need dashboards, because AI can connect to db directly and generate dashboards on-demand
* we don't need utility SaaS, like managing your Google Ads, Meta Ads. Because AI can connect to them directly and manage your ads
* we don't need many other services, because AI can do them
Only moat seems to be in the platforms with large user base and inertia or things run on top of them: Gmail, Google Ads, Reddit, Instagram, MS Office, AWS/Azure/GCP and etc,.
If every product can be made easily, what’s left is relation with the customers. Some companies rely on not just the core code or physical product but rather on the overall ecosystem of customers, service and other things. Here, the alignment between all of them is what matters because for certain products to work, you need all parties to jointly trust in the system. For example what makes Uber work is not only the quality of product but that all parties - drivers, riders all jointly trust that the system works.
A company has moat if they can earn this trust.
He set all ours up from scratch in a week with assistants and he’s never been happier cause he knows exactly what to do and what the outputs should be
He can still teach juniors the important parts and walk through how to use Prometheus and Grafana etc…
You post that claude is just so wonderful and life changing on a forum for nerds and you get some money
hope it helps
for consumer users, yes, it is getting expensive. But for corporations, they are willing to pay the price to be competitive
Depending on how large the paywall ends up being this could lock a lot of those out of being competitive
One must compare to other models such as Geminis ad GPTs.