lambda

Wtf is intelligence!

[lambda @AiKnowsSmth]

@caleb_friesen Yeah they kind a saw this early. Their Manufacturing as a service model is one of a kind.

[lambda @AiKnowsSmth]

@ExaAILabs Data source marketplace for Agents. Sweet.

[Ramchand Kumaresan @Mechramc]

TamilLM needs your help. Many folks reached out from my last post stating willingness to help.
We need real Tamil examples from real Tamil speakers: everyday Tamil, formal Tamil, Tanglish, cultural explanations, literary explanations, regional usage, technical explanations, anything natural and useful.
Goal: 50,000 rows.
But honestly, I will take anything I get. Even 50 good examples from you helps.
Template:
https://t.co/tkMOtQDCRU
How to help:
Open the sheet
Download it or make a copy
Fill 5 to 20 rows if you can
Write in your own words
Send it to [email protected]
Please do not paste copyrighted books, lyrics, movie dialogue, private chats, or scraped web content. If you care about Tamil having a serious open language model, this is one of the most useful ways to help. This will be part of the dataset I push to @huggingface along with the model when done. 🙏

[lambda @AiKnowsSmth]

@random_walker @Zai_org GLM 5.2 with such a UIUX would be great!

[lambda @AiKnowsSmth]

@asmah2107 Is this it!?

[lambda @AiKnowsSmth]

@jaibellare @hackerfabindia Crazy Crazy

[lambda @AiKnowsSmth]

@Arsenal

[Vikas Vij @TheClubJunto]

China Cuts Maximum Degree Programs in IT

China removed 12,000 degrees deemed obsolete in AI era. IT Systems degrees topped the elimination list. Indian education needs an AI pivot. Supply of IT engineers must adjust to demand.

Education Pivot from IT to AI

a. Data from China’s Ministry of Education shows that 12,200 degree programs were aggressively restructured at Chinese universities between 2021 and 2025.

b. Among the degree programs that were targeted, “IT Management and IT Systems” (IMIS) topped the elimination list with 160 nationwide programs entirely cut.

c. The core curriculum of IMIS in China includes IT project management, IT systems design, web/app programming, and data analysis. Graduates typically work as IT Consultants, System Engineers, System Analysts, or IT Strategy Planners.

d. China is deleting this IT major and other associated skills made obsolete by AI. The country is replacing these courses with degrees in new fields like Quantum Computing, Embodied Intelligence, Advanced Robotics, and Semiconductor Engineering.

India’s Current Challenge

a. The exact IT skills that China is removing from its university ecosystem are the ones that still employ millions of Indian IT professionals at the base of the pyramid.

b. However, now a single AI-fluent engineer using English or Mandarin prompts can perform tasks that previously required 5- or 10-person project teams.

c. Millions of mid-level Indian IT professionals with 5 to 12 years of experience currently operate as IT project managers, system administrators, and functional consultants. Their roles are increasingly vulnerable to AI compression.

d. While the demand for Indian IT labour is diminishing, the supply side continues relentlessly with 1.5 million engineering graduates produced annually.

e. The Unstop Talent Report 2025 showed that 83% of engineering graduates in India remain unemployed or without internships. Most of them do not meet the industry standards required in the AI era that would make them employable.

f. India’s IT services sector continues to train and deploy engineers in Application Development and Maintenance (ADM), IT operations management, and enterprise system coordinator rolls (Level 1, 2, and 3 support.) These roles are ripe for disruption.

What India Is Not Doing

a. China’s Ministry of Education eliminated 160 IT degree programs because data showed it was producing graduates that the economy no longer valued. China followed the data; cut the supply; redirected the talent to new fields.

b. India’s policy response to a similar obvious AI challenge at the education level is: inaction. India has no data-driven audit of graduate employment outcomes, and no degree program elimination mechanism.

c. Indian agencies continue to approve colleges and degree courses without any correlation to graduate employment rates. The institutional feedback loop that would listen to market signals does not exist in Indian education.

d. Among India’s IT workforce, those doing coding, testing, maintenance, and L1 support are in the “elimination zone” (in millions, at the base of the pyramid.)

Those performing domain-specialist work are in the “transformation zone.” And those performing AI-native, research-adjacent work are in the “growth zone.”

ENDPIECE

The threat to India's economic future is not AI. It is an education system that continues to prepare students for a world that no longer exists.

@arabicatrader

[lambda @AiKnowsSmth]

@Indianinfoguide & double the salary of teachers and school staff. Let's Goo!!!

[lambda @AiKnowsSmth]

@rohanpaul_ai We need this to be updated QoQ

[Rohan Paul @rohanpaul_ai]

AI agent can get better at long tasks without retraining the agent itself, by using a separate small model to clean and organize its context.

Moves context management outside the agent, so a separate helper can clean up the task history while the main agent stays unchanged.

The paper proposes AdaCoM, which is a separate LLM that edits the agent’s working context before the agent takes its next step.

AdaCoM places a separate, trained manager between the task history and the frozen agent, so the agent does not need to learn a new memory habit or expose its weights.

Before each step, this manager can rewrite, merge, prune, or preserve parts of the running context, then the original agent acts on the cleaned version.

That sounds like summarization, but the distinction matters.

A summary assumes the right answer is compression, while AdaCoM learns that different agents need different kinds of context to stay competent, because stronger agents can use more raw history while weaker agents need shorter and cleaner notes.

They tested AdaCoM on web search and deep research tasks across several agents, and it improved average web search performance by 39%.

----

Link – arxiv. org/abs/2605.30785

Title: "Learning Agent-Compatible Context Management for Long-Horizon Tasks"

[Arsenal @Arsenal]

🔴 𝗧𝗘𝗔𝗠𝙉𝙀𝙒𝙎 ⚪️

Your Champions League final line-up.

COME ON YOU GUNNERS!!

[lambda @AiKnowsSmth]

@bindureddy That's still less than 1% on ARC- AGI 3 lol

[lambda @AiKnowsSmth]

Gemini 3.x has cooked!

[lambda @AiKnowsSmth]

@emollick Hahahaha yea

[HandofArsenal @HandofArsenal]

Greatest picture I have laid eyes on

[lambda @AiKnowsSmth]

@AksharVastarpar @Vicharak_In You can collaborate with IITs, people like @amit_sethi sir can help!?

[lambda @AiKnowsSmth]

@AksharVastarpar @Vicharak_In I think he's the only elected politician to do this... His GitHub is lit!!!

[lambda @AiKnowsSmth]

@sama Brother keep it simple, price/intelligence so that speed becomes irrelevant... Basically zero-shot

[lambda @AiKnowsSmth]

@SynabunAI @ramxcodes 💯