Jamieson O'Reilly
@theonejvo
Hacker. T̶h̶i̶n̶k̶i̶n̶g̶ Doing outside the box.
Founder @d_vuln
Breaking frontier AI models at
Building the worlds most dangerous AI @tryaether_ai
1.3K Following 23.6K Followers
As uncle @JensenHuang would say, that loser premise makes no sense to me.
Capability doesn't advance only as a smooth race for the biggest general model or the most FLOPs. The frontier is jagged, and the cleanest illustration is the strawberry problem.
Ask a model that can hold its own with a physics PhD how many R's are in the word "strawberry" and for a long stretch it would answer two, confidently, every time. The same system could work through a graduate-level proof and then insist 9.11 is a bigger number than 9.9.
Each specific embarrassment gets patched as it goes viral. The shape underneath survives every patch, that is - enormous strength in some places, sudden collapse in others, and no way to tell which ground you're standing on until you test it.
But counting letters and numbers are funny failures.
The expensive version of the same problem shows up in security. Hugging Face was mid-incident, working an agentic attack, and needed a model to review logs. The commercial frontier models refused due to "safety" reasons.
From inside a guardrail, a defender reading attack traffic and an attacker reading attack traffic look identical, so the safest available behaviour is to decline. The team switched to an open-weights model on their own infrastructure and got the analysis done. The most capable system in the room was the one that couldn't help, at the exact hour it was needed.
That refusal is one of many permanent structural feature of general-purpose models, and it sits right on top of a business.
Aether AI operates in the gap between those two failures, and @tryaether_ai AI is Australian.
So telling Australian founders the game is over at the model layer, dig up some lithium and prepare a robot tax, is exactly the wrong message.
The scoreboard was never general-model scale. It's about who owns the jagged edges that customers, regulators, and national security actually pay for. Australia can own plenty of them. Some of us already do.
Show more
When is my first assignment?
This is very true. I know a guy who owned an online casino averaging $10M bets daily - when I showed him his source-code was leaking he basically just laughed it off. It took me a while to understand but as Ernesto points out, distribution and trust are is THE moat. Anyone could have cloned his repo's, far fewer could have contacted his long-term whale punters and convinced them to come over (let alone get the details in the first place).
Show more
Heres how to adapt and survive AI:
from someone making $200k/mo at 22
1. Use your skills to provide a service in the real world.
There are people making $12M per year
with a cleaning company.
Use AI to automate a service business end to end.
A relative of mine makes $140k per month doing this
in a completely different service niche.
2. Build a SaaS which value is not in the code,
but in its network.
A great example of this is @sideshift_app.
Their app is easy to clone,
just like almost any other app is now.
But the network of tens of thousands of creators and companies using their platform is a moat that you cannot clone with Claude.
for example: you can now one-shot Instagram with a single Claude prompt.
But you cannot clone the 1 billion users
that make Instagram worth using.
3. Build a SaaS that requires special access.
Companies like Postiz or Postbridge require permissions that can sometimes take weeks to acquire.
99.9% people would rather pay you
than fill out all of those applications themselves.
4. Build an app or SaaS for consumers, not founders.
Contrary to what @levelsio and many other indie hackers are saying in this thread of doom about the decline in indie hacker revenue,
our apps revenue has actually grown 7x
in the last two months.
We are growing faster than ever
on all fronts.
Consumerism is at an all-time high,
and it will continue to be for years.
The problem is that Levels and many other indie hackers are selling to people on X
who are too smart to buy their middle man apps.
Everyone on X knows they can already do this
with one prompt inside of ChatGPT.
they can probably still thrive if they look for alternative marketing routes outside of X.
5. I can think of 100 more ways to make money with AI
→ Agency selling AI services
→ Personal brand
→ UGC for AI companies that have millions in funding
I mean the list goes on and on.
And im not just telling you this
i've done most of these businesses and made a
ton of money on each.
ultimately my favorite is B2C apps.
because it works the best by far.
don't let this thread scare you into a state of paralisis
And doomer thinking.
levels is speaking from personal experience here.
Apps and SAAS are very much profitable.
The only adapting you have to do,
is being a better operator and not
get stuck on the same channels.
find a solution and keep pushing forward
Show more
Enterprise B2B clients when you ask if they can commit to a multi-year deal up-front.
@theonejvo Right, all the "i reverse engineered" people after realizing this
Did I just unlock claude-fable-5-lite? 😂
Since Fable 5 got pulled (US export control order, Anthropic is contesting it), I wanted to see how much of its character lives in the system prompt vs. the model itself.
I ran the leaked Fable 5 prompt on Opus 4.8 head-to-head against stock 4.8.
Show more
