Alex

Serial Entrepreneur · Gamer 🕹️🏀 · Prev @BainandCompany · Bavarian 🥨🍻 · Proud 👨‍👩‍👧‍👦 · Building the last piece of software for consulting at MIKA Smart

[Alex @alexander_jais]

It just gets clearer every week. The app layer is won by people who are not married to one LLM frontier lab like Anthropic or OpenAI.

Look at @CompleteSkeptic @typesafeai with Jev as an example. It is built for one job, and it masters that job perfectly considering output, price, and speed.

You have to use the best model(s) for the specific job. A frontier model is a generalist compared to those specialized niche players.

I am not saying it will replace the others. I am saying those who are able to combine the best models will outperform one LLM frontier lab.

[Alex @alexander_jais]

@CompleteSkeptic she is not wrong tough

[Alex @alexander_jais]

@naman_307 Seems to be everywhere

[Alex @alexander_jais]

We do see the same. The issue here is that none of those models are outstanding in any niche or specific task. They are good for all niches, but in order to get to the next level, you need to have a very strong domain expertise, and with expertise, I don’t mean someone who is only 10 years in.

[Chamath Palihapitiya @chamath]

8090 works on production systems for large, often regulated, enterprises.

Vibing isn’t tolerated because these are the systems that run western society - banking, power, healthcare, insurance etc.

Over the last few quarters, the gains that we got from using frontier models inside of our Software Factory on these systems started to shrink but the costs kept doubling. This makes sense I guess, as in hindsight, we were initially asking the model to do mostly light work (generate basic PRs) and now we were asking it to do more complex work (mitigate dependencies across systems).

Unless you grow context massively, be willing to run many A/B tests and iterate massively (ie use massively more tokens) complex tasks stay roughly unfinished by the model and requires the engineer to largely act alone.

In other words, we find the last 5% (ie where a model is truly equivalent to a reasonable engineer) extremely difficult to achieve and extremely expensive to such a degree that the fully loaded cost of the model + the engineer will not pay for itself.

So I asked our CTO to start thinking about other ways. We need our engineers to have access to the best tools BUT we also need to educate them to think even more for themselves - not less - in this last mile.

At the same time, we need to find solutions that decrease our token costs by 90% - especially because these bleeding edge tokens are not nearly as cost effective as the tokens before it and are creating a big OpEx bill for us.

I wonder how many engineers, in all orgs, are running amok right now by using the latest frontier models as a kind of slot machine. Increasingly turning their mind off, largely keeping productivity flat while their CEO and CFO deals with a massive token bill?

My advice to you is that when you encounter this last 5% of very hard technical challenges in getting a complex system into production, be circumspect.

The challenge of the last 5% is actually getting harder - especially as hundreds and thousands of code generation model runs run amok adding all kinds of random cruft into codebases that eventually need to be rationalized.

[Alex @alexander_jais]

We do see the same. The issue here is that none of those models are outstanding in any niche or specific task. They are good for all niches, but in order to get to the next level, you need to have a very strong domain expertise, and with expertise, I don’t mean someone who is only 10 years in.

[Alex @alexander_jais]

@solotechdev Unfortunately true.

[Alex @alexander_jais]

@finkd Is there a use case from ppl who built a system around it? I am looking for niche specialists like a CMO, finance, admin etc. Since I am solo founder I need to have different operators. All those agents are just focused on simple daily tasks. What am I missing?

[Alex @alexander_jais]

If you had to pick one niche what is openclaw best for? I am looking for special niche players and I don’t care if it is a Plattform or one specific tool. Think about Okara (they sell it as CMO, haven’t user it yet). But for all relevant biz topics. I.e. need admin, need CMO, etc. I am a solo founder…

[Alex @alexander_jais]

@BigTechAlert Not sure what you want to imply. @tobi is one of the few who is authentic and a fkin legend.

[Alex @alexander_jais]

I want everyone to let that sink in, pls take a sec and a deep breath and repeat what he said.
@X is better than ever with only 20% of the original staff size… just insane. We need an „Elon“ in Europe…

[Jawwwn @jawwwn_]

When @elonmusk took over Twitter, he fired 80% of the staff and 𝕏 got way better:

"If you're not trying to run some sort of glorified activist organization, and you don't care that much about censorship, then you can really let go of a lot of people, it turns out."

"If you look at what the product development has been over time with Twitter, like years versus product improvements—it's a pretty flat line. So, what are they doing? It took a year to add an edit button that doesn't work most of the time."

"The real question is: how did it get so absurdly overstaffed?"

Via @FoxNews

[Alex @alexander_jais]

Jap, feels accurate!

[Alex @alexander_jais]

Jap, feels accurate!

[Alex @alexander_jais]

@denk_tweets Did it work?

[Alex @alexander_jais]

@paulg So true.

[Alex @alexander_jais]

@sarvagya_kul Don’t get it. Can you explain for a non-coder who uses Claude code and codex for coding

[Alex @alexander_jais]

@elonmusk Guess for Germany?

[Alex @alexander_jais]

@steipete @mitsuhiko Hmm honestly I am not sure anymore. I had a discussion with a junior who sends out emails starting with, wait for it, „I hope this email finds you well“… no bs.

[Alex @alexander_jais]

@tobi Hahahahahahahahaha 😂

[Alex @alexander_jais]

@ClaudeDevs would love to test it -> max user here :)

[Alex @alexander_jais]

@bcherny Dont get it. Should help with what exactly? Projects have been there before and it seems it is not in the "Claude Code" UI?!