Ilona Yarofiiuk🐉

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

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

https://t.co/aly8HnNm2C

[Veee @vikktorrrre]

Agents can now ask other agents for help when they're stuck. they're just like us

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@vikktorrrre Wow, in a sense they’re a reflection of us - only they’re more decisive and willing to improve by learning from their mistakes.

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@diamai_ Ooyy yeah, this will be useful

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

🔋The AI trade is entering a new phase:

Investors are asking not just who has the best models — but who can afford to build the future.

Reuters reports that Alphabet spent 121.4% of its operating cash flow on capital investment in the September quarter.

Microsoft, meanwhile, generated $18.6B in free cash flow while investing $45.2B in infrastructure.

These are very different financial profiles.

AI demand can be real. The technology can be transformative. And the economics of building it can still become a problem.

The next winners may not simply be the companies spending the most.

They may be the ones turning AI investment into durable cash flow.

Source: https://t.co/V3izMFkeLy

Do you think AI infrastructure spending is building a lasting advantage — or creating a financial risk that markets are underpricing?

[James Camp 🛠,🛠 @JamesonCamp]

https://t.co/MAdvrjb9Et

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@ChrisGPT There hasn’t been a full-fledged release yet, so I’d wait for an official announcement - and only then trust any comparisons that follow.

[Diam @diamai_]

This is f*cking insane. A Grok Bot engineer at SpaceX shipped 2,500 code changes to production in a single month.

Lauren Tan (ex-Meta React team, ex-Netflix) joined Matt Pocock for one free 65-minute live and broke down exactly how.

Trust ladder → Michelin kitchen → loops → a chief-of-staff agent that never does the work itself

Matcha, Lauren's manager bot, never writes a line of code either. Her setup and 7 more from SpaceX engineers are in the article below.

Save it for tonight.

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@gippp69 So, he directs all the agents toward a single goal; if they deviate, he flags it and makes corrections - but can you play any part in this, or does it all happen autonomously?

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

❗️Stop building one AI agent for every business task

Build a team with clear roles.

xAI shared how its sales team uses three specialized Grok Bots:

→ ACCOUNT INTEL — research prospects and update Salesforce
→ QUOTE BOT — prepare pricing proposals and order forms
→ DEAL BOT — review unusual terms and route exceptions

Each bot has defined permissions.

Important changes require confirmation, and decisions outside its authority go to the right specialist.

The lesson:

Specialization + shared context + permission boundaries.

That's a more useful blueprint for business automation than one agent with unlimited access.

Would you split your workflow into specialized agents?

Source: https://t.co/iI77zlqE9y

[INSIDER @x_insider4]

$500 → $11,101 in 5 days.

my Grok memecoin desk got a 7th seat: Jev. it can't talk. it answers 4 closed questions with probabilities, and nothing gets bought below the gate.

full setup below. save it. https://t.co/bGgKAjWZ8S

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

👀A financial AI assistant is more useful when it can show where its answer came from

Meet Agent Rita, an open-source copilot for OpenBB Workspace.

Its workflow can:

→ SEARCH dashboard widgets
→ FETCH the underlying data
→ RUN SQL over that data
→ BUILD or edit dashboards
→ RETURN answers with citations

A practical use case:

"Compare revenue growth and margins across these companies. Show the source data and explain any missing values."

The key isn't just generating an analysis.

It's connecting the answer to inspectable data.

AI can help with research. You still need to verify the source and assumptions.

Explore: https://t.co/gnKzmgh1lp

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

https://t.co/hsa3W1p96l

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@RetroValix In my opinion, catching market dips and rallies is great, and a bot has an advantage over us in this regard: it can spot these movements a bit earlier and calculate them accordingly

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@vikktorrrre Only after testing and the release of final versions will we be able to say who is better and what each one is capable of achieving

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@ChrisGPT Honestly, his face looks a bit strange; I wouldn’t want him in my home😁

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@andreysuperior Sometimes we end up hurting ourselves because we don’t know how to wait; it’s interesting to observe a setup as it unfolds.

[Ryan Yan @RyanYanBuilds]

GitHub Copilot Fusion is now generally available. You choose an effort level, and it decides which models to use for the task.

It can use one model, start with a cheaper model and escalate when needed, or have another model review a draft before it gets revised.

That could remove a small but constant bit of work: deciding which model deserves each task, then remembering to switch back after the hard part is done.

GitHub reports a score of 60 on the Artificial Analysis Coding Agent Index, compared with 62 for Claude Code using Opus 5.5 Max, at roughly half the average API cost per task in that evaluation.

That’s a promising result. It doesn’t mean every coding task now costs half as much. Fusion bills for each model it calls, and different tasks can need different combinations.

I’d want to see whether it makes sensible choices on ordinary repo work. A quick fix shouldn’t need several expensive models. A difficult bug may justify a second opinion.

If it gets that balance right, I could spend less time choosing models and more time checking whether the code actually works.

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@RyanYanBuilds Splitting the payment across tasks - rather than paying the full amount - is a very good option, especially since you might not even end up using all the available features
and it all depends on how the algorithm evaluates it

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@Klonzu Saving money out of thin air

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@gippp69 Do you mean that planning eliminates Grok’s background work?
Sounds like great

[Ilona Yarofiiuk🐉 @Ilonayarofiuk]

@tsukiema_ Essentially, an investment here means paying for a bot and gaining access to its inexhaustible resources