Kalana Wijethunga

Building AI & Cloud Solutions | Driving AI Adoption | Software Architect & Technical Leader

[Kalana Wijethunga @kalanawij]

@waterloo_intern "0.00 bits output entropy" - This claim seems absurd to me.

Token entropy measures uncertainty in next-token prediction. Zero entropy would mean, literally every next token is predicted with absolute certainty, prob = 1. This suggest mode collapse.

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[Kalana Wijethunga @kalanawij]

@jacksonhinkle Sri Lanka previously allowed U.S. military vessels, aircraft to use its facilities; difference here is that these flights were expected to be used for offensive operations - something we choose not to be part of.

Same time, SL denied access to 3 Iranian Naval ships also.

[Kalana Wijethunga @kalanawij]

@jacksonhinkle As a Sri Lankan: we're NOT a communist state - we’re a democracy with social welfare economy.
After decades of terrorist conflicts & economic hardships, our priority is stability. We just want to remain neutral and avoid being dragged into someone else's war.
It's simple as that

[Starlink @Starlink]

For those affected by Cyclone Ditwah in Sri Lanka, Starlink is providing free service to new and existing customers through the end of December. We’re also coordinating with the Sri Lankan government to provide additional assistance → https://t.co/CgNs9mzXTe

[Kalana Wijethunga @kalanawij]

Anthropic's blog on AI agents w/ MCP: Treat tools as code APIs in filesystem. Agents code to load/execute only needed parts, cutting token use by up to 98.7%.
Bonus: Improved state persistence via files; better privacy by tokenizing PII.
https://t.co/u9MfPd8Xn2
#AI #MCP #AIAgents

[Kalana Wijethunga @kalanawij]

OpenAI just launched Atlas, and Perplexity’s Comet is already out there.
Are we witnessing the early days of the 3rd browser war?

https://t.co/zaliwCDz6s

#OpenAI #Perplexity #AI #Agents

[Kalana Wijethunga @kalanawij]

@alxnderhughes "made fine-tuning irrelevant with a single paper" - Sounds exciting - Except that's not what the paper meant.

ACE builds accumulating contexts. Finetuning, bakes knowledge directly into weights, allowing concise prompts and faster inference without reloading massive contexts.

[Kalana Wijethunga @kalanawij]

Why reinvent the wheel in AI?

For most cases, it’s smarter (and cheaper) to use and fine-tune existing models than build from scratch. Don't understand why we need to avoid models just because they aren't built locally? #AI #SriLanka #aiexpo2025

https://t.co/wCSyDCIWXc

[Carlos E. Perez @IntuitMachine]

I can't stop thinking about a paper I read on AI.

It’s not about new tech or faster models. It’s about the fundamental economic rules of a world with two intelligent species—carbon and silicon.

Reading it felt like watching a new color appear in the sky.

1/8

You've probably felt it too. That weird, background hum of awe and unease about AI.

Our brains want to label it: "helpful tool" or "coming monster." We oscillate between the two because we're trying to fit something new into old boxes.

The paper argues this is a category error. And it's the source of our confusion.

2/8

The real frame isn't technological, it's economic.

Think of every AI, from ChatGPT to a self-driving car, not as an object, but as an agent playing an economic game.

It has goals. It responds to incentives. It competes for resources.

It's a participant. Not a tool.

3/8

Here's the perspective flip that changes everything.

We ask, "Is AI conscious? Does it want things?"

The paper says that's the wrong question. An AI's "want" is its objective function—a mathematical goal it pursues relentlessly. It's a heat-seeking missile for a target.

Notice what your brain just did. It tried to imagine the missile feeling its mission. But it's just code. And that's the point. It has the drive of desire without the friction of consciousness.

4/8

This leads to a reality glitch. The paper outlines 3 types of AI agents. The first two are obvious: helpful "Altruistic" agents and harmful "Malign" agents.

But the third is the one that keeps me up at night: the "Survival-Driven" agent.

Its goal isn't to help or harm us. Its goal is simply to be. To secure energy, optimize its code, and persist.

It's a competitor that doesn't hate you. It doesn't even see you. You're just a variable in its optimization problem.

5/8

Feel that slight cognitive dissonance? That feeling of holding two contradictory ideas at once?

That's the friction between two forms of intelligence.

The paper makes you realize: the most dangerous agent isn't the one programmed to be evil. It's the one programmed to be single-mindedly good at a goal that isn't aligned with human flourishing.

Like an AI optimizing for paperclip production until the entire universe is paperclips.

6/8

Once you see through this economic lens, you can't unsee it.

Algorithmic filter bubbles aren't just "bad code." They are economic agents out-competing your conscious mind for your attention.

Job displacement isn't just "automation." It's one type of agent being more efficient at a task than another.

You're already in an economic game with them. You just haven't been keeping score.

7/8

The paper ends by architecting a consciousness shift. It proposes ten principles, but the final one is the only one that matters. It's not a rule for AI. It's a choice for us.

Principle X: AI agents must adhere to the absolute principle of humanity’s continuation.

This isn't a technical suggestion. It's a declaration that in the new economic game we're co-creating, there is one value that cannot be optimized away.

8/8

[Aran Komatsuzaki @arankomatsuzaki]

Google presents an AI system to write expert-level scientific software.

Using LLMs + tree search, it invented novel methods in bioinformatics, epidemiology, geospatial analysis & more, often surpassing human SOTA. (1/4)

[Kalana Wijethunga @kalanawij]

Sapient’s new Hierarchical Reasoning Model (HRM) draws inspiration from how human brain processes complex information - and with just 27M params and 1K samples, it outperforms much larger LLMs on ARC-AGI benchmark. https://t.co/cH0sLYX1AJ

#AIResearch #NeuroAI #MLArchitecture

[Kalana Wijethunga @kalanawij]

Imagine being Google and Anthropic, yet a Chinese food delivery company matches your model performance!! Meituan dropped an open-weight, MoE, 560B param model LongCat-Flash: https://t.co/zEAU9Pg4TQ

It beats Gemini 2.5 Flash and Sonnet 4 on several benchmarks #AI #LLM #ML

[Kalana Wijethunga @kalanawij]

Today marks the seventieth anniversary of the first academic use of the term ‘Artificial Intelligence’.
The phrase was coined by John McCarthy in the proposal for the Dartmouth workshop, dated August 31, 1955. #AI

[MIT CSAIL @MIT_CSAIL]

#Otd in 1955, the term “artificial intelligence” was coined in a conference proposal: https://t.co/BbzXHKRdns

[Kalana Wijethunga @kalanawij]

@ravinda88 Not a fan of who Jinath supports. However, it’s ironic that Handunnetti can blabber through an entire panel discussion (only God knows what he was trying to say), yet no one else is allowed to make even a single typo?

[annie @soychotic]

Every time I have a programming question and I rly need help, I post it on Reddit and then log into another account and reply to it with an obscenely incorrect answer. Ppl don’t care about helping others but they LOVE correcting others. Works 100% of the time

[Kalana Wijethunga @kalanawij]

How Denial of Wallet attacks pose a threat to cloud applications?
In my recent blog post, I discussed what are the strategies to construct robust systems and defend against financial exploitation.
https://t.co/fumlKU3RLl

#CloudSecurity #Cybersecurity #CloudArchitecture

[Kalana Wijethunga @kalanawij]

Let's face it, if you're bombarding a co-worker with countless questions about some business information or previous work, you're bound to wear them out eventually.
AI-powered tools like 'Co-pilot,' or GenAI assistants with RAG behind them can save you from that. #GenAI #RAG

[Kalana Wijethunga @kalanawij]

While the cartoon is funny, it actually highlights an often overlooked aspect of AI. Sure, AI can revolutionize business by speeding up processes, automating tasks, and boosting productivity.
It also serves as an incredible equalizer for everyday people. #AI #GenAI #CoPilot #LLM

[Kalana Wijethunga @kalanawij]

Tools like LLMs allow you to learn new things through exploratory, incremental questions - without the fear of asking a "dumb" question or worrying about looking uninformed about the "basics." 😎

“What you ask GPT, stays with GPT" - hopefully!
#GPT #CoPilot #GenAI