@roman_in_tech
🇩🇪Account based in
About this account
- Account based in
- Germany
- Connected via
- Germany App Store
Account-level information from X, not a live location or the device used for a specific post.
building multi-agent AI systems for PE due diligence inventing AI-native service firm in private markets co-founder @AxionLab based in LUX
Lux
Joined February 2021
- Tweets1.2K
- Following1.3K
- Followers959
- Likes2.9K
the payoff gap isn't about the models. it's adoption.
big tech rebuilt its org charts and its mental models around AI. everyone else bought licenses and kept the old workflows..
the rest of the market will catch up faster than people think. the people part is the hard part
APOLLO: AI HAS DELIVERED “ZERO PAYOFF” OUTSIDE BIG TECH
Apollo Chief Economist Torsten Slok warns that massive AI spending has yet to deliver measurable productivity gains across corporate America.
Revenue per worker and profit margins are rising for the Magnificent Seven but remain flat for the S&P 493.
Meanwhile, Gartner estimates $3.5 trillion in AI infrastructure spending could face impairment risk by 2030.
The key takeaway: AI investment is booming, but returns outside Big Tech remain elusive.
"just clone the SaaS" math always counts the 5 days of building and skips the 2 years of edge cases.
same in AI services. the demo is a weekend. the hours a human still spends reviewing the output is what clients actually pay to make disappear..
Hey @mattyp @romanugarte_ @ericzakariasson
what's different about Grok Bot @bot for a company vs for one person?
considering it for my startup and curious where it pulls ahead of a Hermes by @NousResearch setup we'd manage ourselves.
We’re hiring sustainability consultants who think like operators, not report factories.
PE / MBB background. Strong commercial judgment. AI-native mindset.
Less formatting boxes. More solving actual problems.
Know someone? Send them this↓
axionlab.org/careers
I need to find a slightly unusual consultant:
- MBB-style training + a few years of experience,
but not married to PowerPoint and old consulting workflows 😅
Someone hungry to use AI, agents, and tech to rethink how consulting work gets done.
where would you post this besides LinkedIn?
Looking for a founding engineer,
who can iterate insanely fast without turning engineering into pure vibing
someone who is
1/ experienced enough to avoid obvious architectural pitfalls and bring real technical judgment
2/ hungry and low-ego enough to use AI-native tools, question old defaults, and rethink how software gets built
dm me
not sure when this happened but both Codex and Claude have become daily consumables for me...
and no, I am not even using frontier models
have you ever felt mentally overwhelmed from working intensively with claude or codex?
this is what I feel almost daily…
yesterday I learned something relevant to this problem: it’s called the Yerkes-Dodson law. performance rises with stimulation, but only up to a point, and after that point the curve bends down. too little stimulation and you’re slow and bored. too much and your brain stops keeping up.
with AI this works differently than with regular work. before, stimulation built up gradually: you wrote code, waited for it to compile, read docs, wrote more. there were natural pauses built in. but now with claude or codex there are almost no pauses left. several tasks running in parallel, a pile of context you have to hold in your head, and you’re making decisions far more often than before. most of that time now goes to judging what AI just did and deciding what happens next.
the exact same thing that makes AI so powerful (parallelism, information density, decision speed) is what pushes you past the peak of the Yerkes-Dodson curve.
building a startup means holding two things that pull against each other: irrational optimism about where this is going, and zero tolerance for lying to yourself about where it actually stands
Roman Evsti retweeted
Hot take: I think it's still important to understand the code that our agents write!
In this mega thread (based on my AIE talk today), I will explain why that's the case, and show some ideas for how to efficiently understand code. Alright, let's dive in. 1/
some companies calling themselves "AI-native services" are consulting firms with an AI copilot installed
to be fair, everyone is still figuring this out, us included.
the tools and the processes are being invented in real time. but the principle underneath is already clear, and it is worth getting right.
for me the difference is simple: watch the org chart when revenue triples.
> AI-powered: headcount triples too, just a little slower. The AI made everyone faster, but it never changed the shape of the business. the work still walks out the door inside people.
> AI-native: revenue triples and the team barely moves, because the next unit of work costs close to nothing to produce. The expert lives in the system, not in every deliverable.
a faster horse is still a horse
the difference between AI-powered and AI-native is the difference between a faster horse and a car: same job, but different physics.
software model never won just because it was using new tech. it won because the next customer costs almost nothing to serve. AI-native services is the first time you can sell expert judgment with almost the same economics.
so it is not about whether you use AI. It is about whether your next customer needs another hire, or almost no one...
hey @usemonologue and @naveennaidu_m
is there any way to export or sync all Monologue notes directly into .md files? put them into git, etc.
use case: I want to automatically feed both notes and transcripts into my hermes agent