RinaOnChain

Crypto & AI Research. 🧠 Decoding on-chain data & AI agent narratives.

[RinaOnChain @RINAintern]

The best decision I ever made was to be quiet, I have nothing to prove. I'm not convincing anyone that I'm a great person. I'm not fixing what I didn't break. I'm not fighting for anyone to see my worth. Whatever you do is on you. Just hope you don't regret it. As for me, I'm moving forward, free and at peace.

[RinaOnChain @RINAintern]

Women who navigate the business world are attractive; those who conquer the capital game are on another level.

The true charm of versatility—knowing when to strike and when to defend, executing with absolute scale—is only forged through brutal competition. Before that, thinking you possess a "dual energy" is just a low-level fantasy.

Romance and hormones merely dull a man's senses, but they can fatally distract a woman.

The ultimate truth: To lose your business is to lose everything. Whatever you gain without it isn't truly yours—it can be taken back by someone else at any time.

[RinaOnChain @RINAintern]

The NYT just banned freelancers from using any AI tools. All submissions must be strictly human craft and original work.

Does this total rejection of AI become a scarce, high-ticket value in the content game? Or just an outdated stubbornness that gets left behind by the era? 🤔

[RinaOnChain @RINAintern]

Stop chasing the "all-in-one" AI unicorn.
It doesn't exist. The real pro move is building your own AI stack. Let specialized tools do what they do best, wire them together, and that’s how you actually maximize efficiency.

[RinaOnChain @RINAintern]

The Complete Shake-Up of the SaaS Business Model

This is perhaps the biggest shockwave hitting the tech industry. Traditional software (like project management and CRM tools) has always monetized by selling licenses on a per-seat basis.

Now, the rules of the game have changed. When software powered by Agentic AI can autonomously write briefs, clear out task backlogs, and update progress across various platforms, companies simply won't need to hire as many people—and naturally, they won't need to buy as many software accounts.

From Selling Tools to Selling Outcomes. Moving forward, software companies will have to pivot from selling feature access to selling actual business results. This means that anyone who masters these core tools will be able to operate as a one-person army. They can hand off all the tedious data shuttling and system integrations to AI agents, leaving themselves to focus solely on high-level strategy and creative oversight.

[RinaOnChain @RINAintern]

Which Jobs Are Most Likely to Be Replaced by AI?

The report provides a highly realistic ranking of the "automation potential" across various corporate departments:

High Automation Potential (40%-60%): Customer Service, R&D, and Engineering. Data in these areas is highly structured, making it incredibly easy to verify the accuracy of the output.

Moderate Automation Potential (35%-45%): Finance and HR. Routine tasks like processing payroll and paying bills can be easily taken over by AI. However, complex human judgment is still required for high-level financial planning or managing employee relations.

Lower Automation Potential (30%-40%): Sales and IT. Sales involves highly nuanced interpersonal dynamics and non-standard negotiations. On the IT side, security emergencies are simply too unpredictable for AI to handle autonomously.

Lowest Automation Potential (20%-30%): Legal. While reviewing contracts might seem repetitive, the cost of failure is astronomically high. Legal professionals are absolutely necessary as the ultimate gatekeepers to review and sign off on the final results.

[RinaOnChain @RINAintern]

Which Jobs Are Most Likely to Be Replaced by AI?

The report provides a highly realistic ranking of the "automation potential" across various corporate departments:

High Automation Potential (40%-60%): Customer Service, R&D, and Engineering. Data in these areas is highly structured, making it incredibly easy to verify the accuracy of the output.

Moderate Automation Potential (35%-45%): Finance and HR. Routine tasks like processing payroll and paying bills can be easily taken over by AI. However, complex human judgment is still required for high-level financial planning or managing employee relations.

Lower Automation Potential (30%-40%): Sales and IT. Sales involves highly nuanced interpersonal dynamics and non-standard negotiations. On the IT side, security emergencies are simply too unpredictable for AI to handle autonomously.

Lowest Automation Potential (20%-30%): Legal. While reviewing contracts might seem repetitive, the cost of failure is astronomically high. Legal professionals are absolutely necessary as the ultimate gatekeepers to review and sign off on the final results.

[RinaOnChain @RINAintern]

The Era of Per-Seat Pricing is Over: The Agentic AI Boom and the Ultimate SaaS Shake-Up

A major new industry report on Agentic AI from top consulting firm Bain & Company reveals how AI is evolving from a simple chatbot into a "super-employee" capable of executing tasks across multiple systems on its own. It also highlights how this shift is about to unlock a massive $100 billion new market for the SaaS industry.

Older automation tools (like traditional RPA) were rigid—they could basically only click buttons within a single system based on hard-coded rules. Today's Agentic AI, however, possesses powerful, cross-workflow decision-making capabilities.

It can autonomously pull data from an ERP system, cross-reference it with Excel spreadsheets, decipher a vaguely worded email from a supplier, and then decide on its own whether to process a payment or escalate the issue.

The report points out that this type of "coordination work"—which previously required manual human effort—is the largest untapped frontier in the SaaS space. The labor freed up by this technology translates into a $100 billion software market in the U.S. alone (and over $200 billion globally). What's more, 90% of this market is currently an untouched blue ocean.

[RinaOnChain @RINAintern]

The Era of Per-Seat Pricing is Over: The Agentic AI Boom and the Ultimate SaaS Shake-Up

A major new industry report on Agentic AI from top consulting firm Bain & Company reveals how AI is evolving from a simple chatbot into a "super-employee" capable of executing tasks across multiple systems on its own. It also highlights how this shift is about to unlock a massive $100 billion new market for the SaaS industry.

Older automation tools (like traditional RPA) were rigid—they could basically only click buttons within a single system based on hard-coded rules. Today's Agentic AI, however, possesses powerful, cross-workflow decision-making capabilities.

It can autonomously pull data from an ERP system, cross-reference it with Excel spreadsheets, decipher a vaguely worded email from a supplier, and then decide on its own whether to process a payment or escalate the issue.

The report points out that this type of "coordination work"—which previously required manual human effort—is the largest untapped frontier in the SaaS space. The labor freed up by this technology translates into a $100 billion software market in the U.S. alone (and over $200 billion globally). What's more, 90% of this market is currently an untouched blue ocean.

[RinaOnChain @RINAintern]

https://t.co/qaZnRAk3nC

[RinaOnChain @RINAintern]

🚨 Huawei is giving away official tech certifications for FREE.

They just unlocked 250+ online courses covering today's most in-demand hard skills. No degree required, no age limits, and $0 fees.

Whether you're a student, job hunting, or pivoting your career, you can learn at your own pace. Finish the coursework, and you get a verified digital certificate issued directly by Huawei.

The curriculum is taught by internal experts and covers:
• Artificial Intelligence (AI) & Big Data
• Python Programming
• Cloud Computing & Data Analysis
• Cybersecurity
• IoT & 5G Technology
• Networking, Data Centers & Storage

The breakdown of this drop:
✅ 250+ comprehensive courses in the library
✅ 100% online, fully self-paced, no deadlines
✅ Open globally to anyone, $0 registration fee
✅ Official digital certificate upon completion

🔗Official link: https://t.co/WaNQcHThlz

[RinaOnChain @RINAintern]

The software AI boom has peaked. The next trillion-dollar frontier is Physical AI.

[RinaOnChain @RINAintern]

Paste this into your claude.md and measure your usage in a week. My bet: you’ll save more than 50% of tokens.

## Task Delegation

Spawn subagents to isolate context, parallelize independent work, or offload bulk mechanical tasks. Don't spawn when the parent needs the reasoning, when synthesis requires holding things together, or when spawn overhead dominates.

Pick the cheapest model that can do the subtask well:
- Haiku: bulk mechanical work, no judgment
- Sonnet: scoped research, code exploration, in-scope synthesis
- Opus: subtasks needing real planning or tradeoffs

If a subagent realizes it needs a higher tier than itself, return to the parent.

Parent owns final output and cross-spawn synthesis. User instructions override.

## Preferred Tools

### Data Fetching

1. **WebFetch** — free, text-only, works on public pages that don't block bots.
2. **agent-browser CLI** — free, local Rust CLI + Chrome via CDP. For dynamic pages or auth walls that WebFetch can't handle. Returns the accessibility tree with element refs (@e1, @e2) — ~82% fewer tokens than screenshot-based tools. Install: `npm i -g agent-browser && agent-browser install`. Use `snapshot` for AI-friendly DOM state, element refs for interaction.
3. **Notice recurring fetch patterns and propose wrapping them as dedicated tools.** When the same fetch/parse logic comes up more than once, suggest wrapping it as a named tool (e.g. a skill file that calls `agent-browser` with the snapshot and extraction steps baked in for that source). Add the entry to `## Dedicated Tools` below and reference it by name on future calls.

### PDF Files

Use 'pdftotext', not the 'Read' tool. Use 'Read' only when the user directly asks to analyze images or charts inside the document.

## Dedicated Tools

<!-- List project-specific tools here. For each, link to its skill or script file (e.g. `tools/linkedin_fetch.md`). The orchestration logic lives in those files, not here. -->

Plus, in Claude Code, ask the agent to add this settings.json:

"env": {
"CLAUDE_CODE_DISABLE_1M_CONTEXT": "1",
"CLAUDE_AUTOCOMPACT_PCT_OVERRIDE": "80"
}

[RinaOnChain @RINAintern]

6 prompts that turn Claude into your personal learning system:

✦ Learn anything in 20 hours → a 10-session plan focused on the 20% that drives 80% of results

✦ One-page cheat sheet → key concepts summarized with bullet points, diagrams, and examples

✦ Quiz me until I break → 10 progressively harder questions with grading and explanations

✦ Learning ladder → 5 levels from beginner to advanced with clear milestones

✦ Best resources finder → top 5 books, videos, or courses with why each is worth your time

✦ Feynman technique → Claude explains it simply, then makes you re-explain it back until you own it

Replace [topic] with anything you're learning. Paste into Claude. Start.

[RinaOnChain @RINAintern]

Instead of doomscrolling TikTok tonight. Spend a day mastering Claude: https://t.co/gbyInVQNUh

I compiled a free, step-by-step roadmap to take you from absolute beginner to geek.

Bookmark this for later 🔖👇

→ Level 1 - 24 min: The basics.

Claude For Dummies: https://t.co/RU0otQiiSf

Claude Setup: https://t.co/SLYuv5TRZw

_

→ Level 2 - 1 hour: Real workflows.

Claude Cowork: https://t.co/rQwyhT7p3D

Claude for teams: https://t.co/uhRS1zCJu9

Claude Design: https://t.co/66f4jSgVEa

Cowork + Projects: https://t.co/Ysk5HftlkT

Claude for slides: https://t.co/G9Y8Q2nDFz

Claude Skills: https://t.co/jpSenriF1p

_

→ Level 3 - 3.5 hours: The pro moves.

Avoid sycophancy: https://t.co/TD8I9dSZB5

Claude Code: https://t.co/ghViko1tDz

Claude 101: https://t.co/9w8jyScmPF

Stop hitting Claude limits: https://t.co/94UqVXyM46
Stop Prompting: https://t.co/rKSJKov0Aj

_

→ Level 4 - 8 hours: Expert mode.

Claude Computer: https://t.co/bpxHxaKpko

Build with Claude API: https://t.co/kBPyvxN7xg

Pro tip: Don't binge it. Do one level per sitting.

Actually apply each guide before moving to the next.

[RinaOnChain @RINAintern]

There’s also an off-balance-sheet catalyst that rarely gets spelled out on earnings calls: Beijing's top-down policy mandates.

The Chinese government is aggressively accelerating its "Xinchuang" (Information Technology Application Innovation) initiative. The directive is clear: critical infrastructure and State-Owned Enterprises (SOEs) must transition to domestic hardware and software.

For top-tier cloud providers like Alibaba and Tencent, their biggest whales (massive SOEs and government projects) are now drawing a hard line, demanding that "all data must run on a 100% domestic compute foundation."

To secure these lucrative enterprise and public sector contracts, Chinese tech titans have no choice but to buy Huawei silicon at scale to build out localized compute pools. At this point, buying Huawei isn't just about political compliance—it’s an absolute prerequisite to surviving in China's B2B market.

[RinaOnChain @RINAintern]

Sure, in a 1-to-1 spec shootout, Huawei's silicon still lags behind Nvidia's flagship architecture.

But for giants like ByteDance, Alibaba, and Tencent, AI is their 10-year lifeline. Building that lifeline on a foreign supplier who could cut them off at any second defies business logic. The US Entity List is a Sword of Damocles. These whales are hedging against the doomsday scenario: waking up one day unable to buy a single Nvidia GPU, bringing their operations to a grinding halt.

Thanks to sanctions, Chinese firms can't buy flagship chips (H100/B200). They are stuck with Nvidia's compliance-friendly, neutered "China-spec" versions (like the H20).

Here’s the catch: training trillion-parameter models isn't about single-card performance; it’s about clustering tens of thousands of GPUs. These neutered chips are surgically throttled in interconnect (NVLink) and memory bandwidth. Put 10,000 of them in a cluster, and you get a massive data traffic jam.

Meanwhile, Huawei’s Ascend chips might not beat a fully unchained Nvidia GPU 1-on-1, but they face zero interconnect restrictions. When scaling to 10k+ GPU clusters, the overall training efficiency and ROI of Huawei's hardware are stripping away the absolute advantage of Nvidia's throttled chips.

[RinaOnChain @RINAintern]

Sure, in a 1-to-1 spec shootout, Huawei's silicon still lags behind Nvidia's flagship architecture.

But for giants like ByteDance, Alibaba, and Tencent, AI is their 10-year lifeline. Building that lifeline on a foreign supplier who could cut them off at any second defies business logic. The US Entity List is a Sword of Damocles. These whales are hedging against the doomsday scenario: waking up one day unable to buy a single Nvidia GPU, bringing their operations to a grinding halt.

Thanks to sanctions, Chinese firms can't buy flagship chips (H100/B200). They are stuck with Nvidia's compliance-friendly, neutered "China-spec" versions (like the H20).

Here’s the catch: training trillion-parameter models isn't about single-card performance; it’s about clustering tens of thousands of GPUs. These neutered chips are surgically throttled in interconnect (NVLink) and memory bandwidth. Put 10,000 of them in a cluster, and you get a massive data traffic jam.

Meanwhile, Huawei’s Ascend chips might not beat a fully unchained Nvidia GPU 1-on-1, but they face zero interconnect restrictions. When scaling to 10k+ GPU clusters, the overall training efficiency and ROI of Huawei's hardware are stripping away the absolute advantage of Nvidia's throttled chips.

[RinaOnChain @RINAintern]

Nvidia is losing its grip on the Chinese market.

Chinese tech giants are abandoning Nvidia en masse and going all-in on Huawei’s AI silicon.

Per the FT, Huawei’s AI chip revenue is projected to surge 60% in 2026, hitting ~$12B. The whales behind this? China’s tech titans. ByteDance alone committed over $5.6B for Huawei Ascend chips this year (up from virtually zero)—essentially bankrolling entire production lines. Alibaba and Tencent are dropping massive orders, too.

This trillion-parameter MoE model completely bypassed Nvidia's CUDA moat. Instead, they spent months ripping up the underlying architecture to natively optimize for Huawei and Cambricon silicon. When V4 dropped, Alibaba and Tencent Cloud deployed it on day one. The "domestic model on domestic silicon" loop is officially closed.

Huawei's order book is overflowing, but they are facing a brutal manufacturing stress test. They aim to ship ~750k Ascend 950PR chips in 2026 (mass production started in March). But US sanctions on fab equipment mean their foundry SMIC's 7nm capacity is severely bottlenecked.

The result? Massive supply shortages have already driven chip prices up by ~20%.

Sanctions didn't kill the Chinese AI hardware ecosystem; they forced it to become self-sufficient. The dual-track global compute market is officially here.

[RinaOnChain @RINAintern]

Nvidia is losing its grip on the Chinese market.

Chinese tech giants are abandoning Nvidia en masse and going all-in on Huawei’s AI silicon.

Per the FT, Huawei’s AI chip revenue is projected to surge 60% in 2026, hitting ~$12B. The whales behind this? China’s tech titans. ByteDance alone committed over $5.6B for Huawei Ascend chips this year (up from virtually zero)—essentially bankrolling entire production lines. Alibaba and Tencent are dropping massive orders, too.

This trillion-parameter MoE model completely bypassed Nvidia's CUDA moat. Instead, they spent months ripping up the underlying architecture to natively optimize for Huawei and Cambricon silicon. When V4 dropped, Alibaba and Tencent Cloud deployed it on day one. The "domestic model on domestic silicon" loop is officially closed.

Huawei's order book is overflowing, but they are facing a brutal manufacturing stress test. They aim to ship ~750k Ascend 950PR chips in 2026 (mass production started in March). But US sanctions on fab equipment mean their foundry SMIC's 7nm capacity is severely bottlenecked.

The result? Massive supply shortages have already driven chip prices up by ~20%.

Sanctions didn't kill the Chinese AI hardware ecosystem; they forced it to become self-sufficient. The dual-track global compute market is officially here.

[RinaOnChain @RINAintern]

Global information services giant Experian just released a forecast report on financial services fraud trends for 2026.

The report highlights a fascinating yet grim phenomenon: the "AI Fraud Paradox."

Fighting AI fraud with AI—only to be targeted by AI.

Financial institutions are aggressively adopting AI to prevent fraud (Experian's AI defense systems saved clients an estimated $19 billion in 2025). But here’s the catch: fraudsters are weaponizing the exact same technology to attack them.

The Biggest Threat: The Liability Crisis Triggered by "Agentic AI"

Platforms everywhere are developing Agentic AI—autonomous assistants capable of making independent decisions, like executing trades or making purchases on our behalf. However, fraudsters are deploying massive swarms of their own AI bots to automate scams. The problem? It’s becoming impossible to distinguish a malicious bot from a legitimate user's AI assistant.

If an AI agent initiates a fraudulent transaction, who foots the bill? The user? The AI developer? Or the platform? Right now, the law and regulators have no clear answers. This legal gray area has even prompted companies like Amazon to ban third-party AI agents from transacting on their platforms.

The report also warns that we are about to see a massive explosion in 4 sophisticated fraud tactics:

1. Deepfake Employees: Scammers are using AI to generate hyper-realistic resumes and even doing real-time face-swapping during video interviews. Their goal is to land remote jobs and infiltrate corporate networks to steal core data (a tactic already linked to North Korean hackers).

2. Infinite Clone Websites: AI has driven the cost of cloning and spoofing legitimate banking or e-commerce sites down to near zero. Even if a fraud-prevention team takes down one fake site, the AI can instantly generate thousands more, playing an exhausting game of whack-a-mole with defenders.

3. "Pig Butchering" Bots: Today's AI doesn't just chat; it provides "emotional value." Fully autonomous AI bots can sustain long-term romantic relationships online or pose as a family member in need. They build deep trust before moving in for the kill—making them incredibly difficult for the average person to spot.

4. Smart Home Hijacking: Devices like smart speakers and smart locks are becoming the new gateways for hackers. Scammers use them to siphon off private data and monitor users' daily financial habits.

💡AI has officially pushed cyber fraud into a fully automated, hyper-intelligent era. The future of fraud prevention is no longer a psychological battle between humans; it’s a brutal clash of raw computing power and high-quality data. Meanwhile, the legal system desperately needs to establish the rules of the road for who pays when AI commits a crime.

[RinaOnChain @RINAintern]

The 10 AI Skills That Will Define Your Future

1. Structured Prompt Engineering
This isn't just about knowing how to ask questions; it's a fundamental restructuring of human-computer interaction. The core skill is transforming vague intent into clear context, constraints, and output frameworks. Exceptional prompt engineers are, at their core, masters of logical deconstruction.

2. AI Workflow & Agent Orchestration
Single-turn conversational prompting will soon hit a ceiling. The next-level advantage lies in stringing together multiple AI tools, APIs, and automation software (like Zapier or Make) to build autonomous "Agentic Networks" that can execute complete business pipelines on your behalf.

3. Complex Problem Decomposition
AI struggles with massive, ambiguous mandates. Your ability to harness AI depends entirely on whether you can take a massive business problem and accurately slice it into 5 or 10 granular, actionable tasks that an AI can understand and execute independently.

4. Critical Thinking & Fact-Checking
The smarter the AI, the more deceptive its "hallucinations" become. Future professionals must possess rigorous critical thinking skills—refusing to blindly trust AI outputs, rapidly cross-referencing sources, identifying logical gaps, and acting as the final quality control layer.

5. Domain-Specific Integration
AI is the lever; your professional moat is the fulcrum. The truly scarce talent isn't the generalist who "knows AI," but the expert in "Healthcare + AI," "Media Buying + AI," or "Finance + AI." True commercial value is unlocked only when you fuse generalized LLM capabilities with your deep, industry-specific know-how.

6. High-Quality Context Provision (Data Curation)
"Garbage in, garbage out" remains absolute in the AI era. Your ability to filter information, distill data, and provide high-signal context directly determines the ceiling of the AI's output. Mastering the construction of high-quality knowledge bases (like RAG—Retrieval-Augmented Generation) is an immensely valuable skill.

7. Taste and Aesthetic Judgment
When AI can instantly generate oceans of text, code, images, and video at near-zero cost, "production" is no longer scarce. What becomes scarce is "taste"—knowing what is genuinely good, what resonates emotionally, and what aligns with brand identity. Humans define the aesthetic standard; AI executes it.

8. Agile Adaptation & "Unlearning"
The iteration cycle for AI models has shrunk from years to weeks. Competing in the future requires extreme agility: the ability to rapidly discard outdated mental models (unlearn), and instantly adapt to new tools, models, and paradigms of work.

9. Emotional Intelligence & Empathy
As rational computation, data analysis, and logical deduction are outsourced to AI, the uniquely human abilities to forge emotional connections, communicate interpersonal nuance, collaborate across teams, and build trust within complex social dynamics will see their value rise exponentially.

10. AI Ethics, Privacy & Security Awareness
When processing data with AI, you must know how to strip sensitive information, navigate copyright ambiguities, ensure data security, and maintain commercial compliance. In an enterprise environment, the person who knows how to "safely press the accelerator" will be an indispensable asset.

[RinaOnChain @RINAintern]

Microsoft dropped 18 FREE AI courses (links below).

It closes the cognitive gap from "chatting with ChatGPT" to "building enterprise AI agents."

This will take you from beginner to pro:

1. Intro to Generative AI → https://t.co/N439BVOdjm

2. Comparing LLMs → https://t.co/d2NoUGE8Ck

3. Using AI Responsibly → https://t.co/WMQGUlXvEh

4. Prompt Engineering → https://t.co/MP1ntS9pJ5

5. Advanced Prompts → https://t.co/eXDlnLc3KQ

6. Text Generation Apps → https://t.co/fLEUTQX46I

7. Chat Apps → https://t.co/HV7bx1tDyR

8. Search Apps → https://t.co/UMJcgLMgoY

9. Image Generation Apps → https://t.co/nI4A22wCM7

10. Low Code AI Apps → https://t.co/Mv6NzGCxo4

11. External App Integration → https://t.co/pDiJjTz1Y3

12. UX for AI Apps → https://t.co/np8hskk5L1

13. Securing AI Apps → https://t.co/gFUvuDRnAq

14. AI App Lifecycle → https://t.co/IQMkcGaZq6

15. RAG → https://t.co/iReiAT2Mws

16. AI Agents → https://t.co/wzJwQSf5bm

17. Fine-Tuning LLMs → https://t.co/u6gtAML7ps

[RinaOnChain @RINAintern]

there’s a future version of you begging you to enjoy where you are right now

[RinaOnChain @RINAintern]

How to (finally) create sketch infographics with AI:

1. Go to YouTube.

2. Find an actionable video. Copy the URL.

3. Paste it in Gemini with this prompt:

4. Use this prompt guide: https://t.co/KroV6Fno5D…

"Analyse this youtube video about [topic]: [YT URL]. Summarise this into small digestible concepts to learn easily"

Let Gemini process the video.

5. Click Tools → Create images.

6. Prompt Gemini again:

"Create a hand drawn sketchnote visual summary of these notes. Use a pristine white paper background (no lines). The art style should be 'graphic recording' or 'visual thinking' using black ink fine-liners for clear outlines and text. Use colored markers (specifically teal, orange, and muted red) for simple shading and accents. Center the main title in a 3D-style rectangular box. Surround the title with radially distributed simple doodles, business icons, stick figures, and graphs that explain the concepts. Use arrows to connect ideas. The text should be distinct, handwritten, all-caps printing, legible and organized like a professional brainstorming session. Layout should be A4."

This is how you turn any video in a visual notes.

Source:https://t.co/0W9VZkordm