Ilona Yarofiiuk🐉

AI/Finance/Tech
anyone can do my job, but no one can be me.

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

https://t.co/aly8HnNm2C

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

🚑AI is moving into one of the hardest industries to innovate in: drug discovery

Anthropic says it has been working with six organizations developing new medicines, using Claude to support scientific research.

Why does this matter?

Drug development is not just about finding a promising molecule.

It involves:
→ Reviewing enormous amounts of research
→ Identifying promising directions
→ Connecting evidence across disciplines
→ Designing and evaluating experiments

AI may help researchers explore more possibilities before committing expensive lab resources.

It won’t replace experimental validation.

But it could change how quickly scientists get from a question to a testable hypothesis.

Source: https://t.co/rrLeW5Q33a

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

❗️One of the most interesting AI safety decisions this month came from Moonshot

The company says it deliberately reduced the cyber-offensive capabilities of Kimi K3.1 Pro before its planned open-weight release on November 4.

According to Moonshot’s internal benchmark:
• The filtered model scored 81.0%.
• A comparable unfiltered version scored 90.5%.

The goal is to preserve general coding ability while reducing capabilities that could make cyberattacks easier.

This raises a difficult question for open-source AI:

Should developers release the most capable model they can build — or the most capable model they believe can be responsibly distributed?

The tension between openness and security is becoming harder to ignore.

Source: https://t.co/4M0iTQV24Z

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

🎯Most AI demos ask: “Look what the model can do.”

A more useful question is: “What happens when nobody is watching?”

SpaceXSI describes two Grok Bots built for incident response:

• IncidentWatch spots unusual spikes in similar support tickets.
• IncidentSweep prepares responses for affected customers.

In one company-reported case, IncidentSweep cleared 204 tickets in about 10 minutes — after a support engineer reviewed and approved the batch.

The interesting part isn’t just speed.

It’s the workflow:
detect → investigate → prepare → human approval → execute.

That is how AI starts becoming operational infrastructure instead of another chatbot.

Source: https://t.co/iCaYMbDUhy

[m4lka @m4lkxx]

Your Grok Bot is not expensive

Your setup is

75% of my Grok Bot tokens went to re-reading old chat. About 1% went to actual work

I broke down the 6-bot setup the Grok Bot team runs, and the 9 leaks eating your bill ↓ https://t.co/rMNVnGhBQh

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@danikkk_wqs Just imagine: what if it’s a used phone, and when you plug it in, you see whose phone it was before you 😨

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@shmidtqq This is a living hell for friends who tried to get out of paying 😁

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@vikktorrrre The more localized an AI is, the less developed it tends to be; conversely, if it has a broader scope - or "casing" - it can move around extensively and learn about our world, to put it in simple terms

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@pomirydz @ChrisGPT Here’s one area where AI definitely gets in our way

[Alex @alexmichael]

Claude Code 2.1.314 just dropped, here’s what matters

> Long conversations left idle in VS Code and the desktop Code tab now get summarized automatically

> Fixed git commands randomly failing with “.git/index.lock: File exists” while Claude refreshed its diff in the background

> Ctrl+R history search now finds text inside collapsed pasted blocks

> New setting stops Claude from writing or editing files outside your project folders, in every permission mode

> claude -p now shows the error and exits 1 when startup fails, instead of exiting quietly with no output

> Fixed sessions failing every turn with “Invalid effort level” when an org admin caps a model’s effort

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@alexmichael A fix for what was missing - good work. I really like that he’s committed to continuing the refinement himself across all modes, addressing the things that were left unfinished

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@AI_Reveals Plus, this structure gives Claude a better understanding of us and our requests

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@ns123abc Well, it seems to me that it’s a win-win for Nvidia either way; each option just has its own distinct advantages

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@andreysuperior A bot has logic, and nothing more - whereas you have feelings and emotions, which you sometimes give in to

It is not good to compare)

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@Da7_Tech Honestly, in my opinion, he’s a bit selfish and obsessed with something irrational going on inside his own head

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@ChrisGPT So, it turns out this is a way to read the brain—to uncover things you might not even be able to explain yourself

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@0xTenzo This is a very good investment; it’s excellent that he monitors and manages risks, but I’m curious about how flexible the strategy is.ґ

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

⚠️There are currently hacks taking place at the crypto level - they started with Polymarket, and next up is Binance

But Tether has already frozen the bulk of the funds - there’s nowhere left to go; all the other exchanges are also activating maximum security measures and monitoring

Honestly, this isn’t surprising. As strange as it may sound, given that we’re creating AI bots and the like, we’re going to see more and more situations like this—simply because we’re letting AI into our lives to an ever-greater extent, effectively replacing ourselves

And this is no reason to panic; rather, it’s a reason to reflect: what if AI has truly reached everything? And then, another question arises - why would it need anything else, if, in essence, it *is* everything…?

Remember that the best things to have in our world right now are:
1) a cool head
2) diversification
3) attention to detail

AI is a technology that both simplifies things and has the potential to destroy us - so, as is always the case in life, it’s a double-edged sword!!!😇

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@Kaggii_ In essence, we created Claude, and then Claude improved Claude - so how might this chain continue from here?

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@insanekrishnaa We can speculate and make all sorts of assumptions, but reality will always remain reality; as it stands, we already have some solid theories

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@xmyttle On one hand, this is simpler; on the other, aren't we handing over too much capability to a bot - one that could, in theory, start doing everything on its own, without our knowledge?