> be cursor
> fine tune kimi2.5/2.6 using user steering data (basically, user tells cursor "don't do this, do that instead")
> build composer (fine tuned LLM)
> do bare minimum LLM research for coding
> build a banger UI/UX setup for cursor based on VSCode
> make shit tonnes in valuation against a teeny tiny team of engineers and sales folk
> get absorbed by SpaceX (val: 60Bn$)

Now, why did SpaceX not make anything? Cuz they suck ass at LLMs and stuck to porn with grok. But they have compute big enough to support Anthropic's workload - now extending to use rich user data with Cursor.

The only lesson here is building stuff on top of open source stuff is probably the best way to find an entry and exit in this slop market. And selling shovels ofc.

https://www.moneycontrol.com/news/business/startup/kunal-shah-to-become-global-ceo-of-whatsapp-meta-invests-nearly-1-billion-in-cred-13955883.html

Indian data sold at cost to Meta by Kunal Shah

WhatsApp Premium Pro (satire)

Meta has the distribution already sorted through Whatsapp, and they've been trying hard to make people use their WhatsApp Pay (probably trying to pull a WeChat in india) - and now with Cred's UI/UX psyops, they'll be fishing for a larger fish.

As adbhim said, the matchup with Telegram being banned and this shit now is just diabolical.

Cash + Signal/Threema perhaps should do

https://x.com/i/status/2071108644838736261

anything to ban telegram

> telegram should be replaced by whatsapp
> UPI payments should get integrated into whatsapp
> tiktok banned around the same time reels became popular on instagram


why is meta everywhere in India?

https://yunwu.ai/

The chinese are reselling subscription plans from American AI Giants at a huge cut in cost by routing models separately 😭

The good never hate on open-source free software.

https://x.com/i/status/2071330294452666695

A new LLM GLM-5.2, while hyped a lot online, is at the "getting there" stage, with little refinements needed. But of course, Anthropic will never be able to take side with such LLMs being available.

Still, these aren't LLMs that a normal user can run on their consumer grade systems, but it allows new smaller enterprises to be able to run their own weights of these LLMs, with complete control of input and output, ensuring their own privacy too.

these orgs are moving to self hosting FOSS LLMs

its simply a matter of control and profitability

don't mistake any of these orgs as privacy respecting - they simply want to cut costs and maintain company intellectual-property

anything to deanonymize

AI faitgue is not real
AI faitgue is not real
AI faitgue is not real
AI faitgue is not real
AI faitgue is not real
AI faitgue is not real

Consuming Cursor has super powered Grok 4.5 now.

Explains the 60Bn $ sale rate.

Gov moves one step closer to mass deanonymization:
https://www.yahoo.com/news/world/articles/shocking-indian-court-case-could-184551656.html

grifting is our way of life

Apple interest thrusts China’s CXMT into memory chip spotlight
https://www.ft.com/content/f4ac5c92-03be-4499-b16a-017a7e9ee228

https://archive.vn/9Cz78

#tech #economics

https://x.com/i/status/2078133895766114412

Always trust the corporate to be evil

Just so everyone knows, Azure and other cloud services run AV services to scan files systems
This can result in downtime/failover/hung state for application hosted

Solution is to exclude those directories from scanning