Jukan
I picked up a small amount of Ibiden shares today.

I agreed with Bernstein’s view that even if Intel wins EMIB-T orders from external customers, it’s actually Ibiden—the one supplying the substrates to Intel—that stands to benefit far more. On top of that, I believe ABF substrates are going to face severe shortages going forward.

And coincidentally, Goldman Sachs just put out a report saying that ABF substrate supply is projected to fall short by a whopping 42% over the next two years.

I’m really disappointed that I suddenly ran short on cash and couldn’t buy more Ibiden.
tweet

God of Prompt
RT @godofprompt: ChatGPT's custom instructions feature is insanely powerful.

But 99% of people write garbage instructions.

I tested 200+ custom instruction sets.

These 5 patterns increased output quality by 3.4x: https://t.co/xVEOxr2Gm8
tweet

Javier Blas
CHART OF THE DAY: By 2050, the world will be consuming lots of oil — or almost nothing.

The annual @IEF_Dialogue review of all the major energy scenarios shows that the high and low points for oil demand in 20250 are apart by ~119m b/d, more than today's consumption. Go figure! https://t.co/nfDZAcYEKc
tweet

God of Prompt
RT @godofprompt: https://t.co/oNpZYTk7a2
tweet

God of Prompt
RT @free_ai_guides: The LinkedIn content grind is exhausting.

You post. nothing happens. you post again. still nothing. you run out of ideas by week 2.

So I made a mega-prompt that generates a full month of content in one shot:

→ researches your audience's desires, fears, frustrations
→ creates 10 posts per week
→ balances growth, knowledge, and authority content
→ maps every post to a specific pain point

One prompt. 30 days of content. Done.

Comment "LinkedIn" and I'll DM it to you.
tweet

App Economy Insights
RT @EconomyApp: $AMD AMD Q4 FY25:

• Revenue +34% Y/Y to $10.3B ($0.6B beat).
• Operating margin 17% (+6pp Y/Y).
• Non-GAAP EPS $1.53 ($0.21 beat).
• Q1 FY26 rev guide ~$9.8B ($0.4B beat). https://t.co/afpWaIWa1j
tweet

God of Prompt
AI benchmarks are lying to you.

Models are scoring 95%+ on tests because the test questions were IN their training data.

Scale AI published proof in May 2024.

We have no idea how smart these models actually are.

Here's the contamination problem nobody's fixing: https://t.co/uPOW7YQyNX
tweet

God of Prompt
RT @godofprompt: AI wrote a billion lines of code in 2025.

And somehow nobody built the one that actually matters for families.

This is the first ever launch of The World’s First Family AI... Nori.

Most optimize work. None optimize life.

Nori is my family CEO.

It remembers soccer practice, plans dinner, and keeps my family functioning 👇
https://t.co/JnoKkAVJwZ
tweet

Illiquid
Ibiden and Semicap are less painful ways of being exposed to this compared to being an Intel shareholder?

At today's Cisco's conference, Lip Bu mentions that if observers see Intel "... putting money into glass substrate... or adding certain capital equipment to scale, it would indicate that 'real customers' have committed."

$INTC spent $1B+ on R&D and built 600+ patents over a decade of research on glass substrate. It is now one of the most important aspects in meeting the demands of AI infrastructure.

Intel is now licensing the IP to Samsung Electro-Mechanics, Absolics, and others. - Ben
tweet

The Transcript
When you see the drawdown on your saas stock: https://t.co/lSDQfX4bWs
tweet

God of Prompt
RT @rryssf_: anthropic just published research claiming ai failures will look more like "industrial accidents" than coherent pursuit of wrong goals.

i read the paper and the original 2023 theory it builds on.

the limitations section isn't where the problems are.

the problems are in the framing itself.

here's what it doesn't mention 👇
tweet