Ajithdhev

Singular Dreamer

[Ajithdhev @ajithdhev]

“My sky is different from your sky”

[UNICEF @UNICEF]

Yes, those photos you’ve seen are real.

More than 18 years ago, a baby named Lamine Yamal and his mom Sheila met Lionel Messi at a UNICEF fundraising photoshoot.

Today, their achievements on the pitch inspire millions. Off the pitch, both Messi and Lamine Yamal use their voices and platforms as UNICEF Goodwill Ambassadors to support and advocate for children around the world.

The goal? That every child survives, thrives, and fulfils their potential.

We are proud to have them on our team.

Credit: Joan Monfort

[Novak Djokovic @DjokerNole]

My longest Wimbledon match ever. An unforgettable nightshift 💪

[Gemini Notebook @Gemini_Notebook]

Doom scrolling but make it educational 🤓

Introducing Short Video Overviews in NotebookLM! Turn your most complex sources into 60-second, vertical videos that deep dive into any concept.

Rolling out now to Google AI Ultra and Pro subscribers on mobile & web (free users soon!)

[Ajithdhev @ajithdhev]

An interesting read.

[Satya Nadella @satyanadella]

https://t.co/vLmiBKTtX3

[FIFA @FIFAcom]

The reason Japan fans clean the stadium after each game. Respect. 🤝🇯🇵

[Anthropic @AnthropicAI]

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.

It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.
https://t.co/NQ7IfEtYk7

[Ajithdhev @ajithdhev]

A beautiful view of downtown Toronto from CN Tower at dusk! #Canada

[Ajithdhev @ajithdhev]

@NASA @NASAArtemis We heard about the early troubleshooting with the Orion’s Amine based CO2 scrubbers and the waste system. As 'test pilots' for the first deep space habitat in decades, what’s the most surprising 'low-tech' solution or tool you’ve had to use to fix a high-tech problem so far?

[Ajithdhev @ajithdhev]

The world just changed. On April 1, 2026, humanity officially ended its 53-year "stay-at-home" phase. 🚀

For the first time since 1972, there are human beings currently hurtling toward the Moon.

Here is the breakdown of why Artemis II mission is a technical marvel:

On Day 6, the crew will reach 252,000 miles from Earth, the farthest any human has ever traveled.

Space "Parallel Parking": Pilot Victor Glover finished a 70-minute Proximity Operations demo, manually flying the 27-ton Orion capsule just 30 feet from a spent rocket stage. This proves humans can take the wheel if docking computers glitch.

Laser Internet: They are testing the O2O system, using infrared lasers to beam 4K video at 260 Mbps. That’s "home Wi-Fi" speeds from the lunar vicinity.

The Mach 32 Homecoming: On April 10, the crew will hit the atmosphere at 25,000 mph. The heat shield must endure temperatures of 2,760°C (5,000°F), literally half the temperature of the Sun’s surface, to bring the crew safely home to San Diego.

The Moon was our finish line in 1969, in 2026, it’s officially our starting block for the stars. 🛰️✨

[OpenAI @OpenAI]

Today we’re introducing GDPval, a new evaluation that measures AI on real-world, economically valuable tasks.

Evals ground progress in evidence instead of speculation and help track how AI improves at the kind of work that matters most.
https://t.co/uKPPDldVNS

[Ajithdhev @ajithdhev]

Building on this is going to be exciting as, in the future, we will get to understand how/what other species communicate! Imagine transcribing a dog barking into natural language!

[Sundar Pichai @sundarpichai]

Introducing DolphinGemma, an LLM fine-tuned on many years of dolphin sound data 🐬 to help advance scientific discovery. We collaborated with @dolphinproject to train a model that learns vocal patterns to predict what sound they might make next. It’s small enough (~400M params) to run directly on Pixel 9 phones used in the ocean! A very cool step toward enabling interspecies communication.

[Ethan Mollick @emollick]

The past 18 months have seen the most rapid change in human written communication ever

By. September 2024, 18% of financial consumer complaints, 24% of press releases, 15% of job postings & 14% of UN press releases showed signs of LLM writing. And the method undercounts true use

[Ajithdhev @ajithdhev]

Grok 3 outperformed OpenAI's GPT-4o, Claude 3.5, Deepseek V3, and Gemini 2 Pro on several benchmarks. Great work by the @xai team!

[Ajithdhev @ajithdhev]

I am a huge fan of the All-In podcast and in their latest episode with Naval, they discussed web crawling, licensing, and copyrights among other topics. A few months back, I built a tool that gathers information from any news article to give a well-rounded view on any topic. So, I built a plugin that can basically crawl into any news article (and break any paywall) extract the content (even from a video), and then summarise everything so you can understand a topic from the ground up from a first principles approach.

Initially, I wanted to make it open source for everyone but I never published it (and probably never will) because I have always believed that good content merits subscription to help support quality journalism and encourage unbiased reporting.

At the same time, if I can build this over a weekend, anyone can do it. So it is super important for news media organisations to better protect their data. Their revenue models are going to change in a big way (it is already changing) so publishers need stronger guardrails. Here's a sneak peek of what it can do.

If you're curious and want to test it out on a link or topic, just let me know, I’d be happy to run it and share the results!

@theallinpod @Jason @chamath @friedberg @DavidSacks @naval

[Ajithdhev @ajithdhev]

PM Trudeau announces 25% tariffs on $155 billion of US goods

[Ajithdhev @ajithdhev]

US 25% tariffs on Canadian goods could spike prices, slash jobs, and weaken the loonie. When steep tariffs are imposed, everyday costs rise as imported goods become more expensive—squeezing family budgets and putting small businesses under pressure.

But the real danger isn’t just higher prices—it’s how Canada might respond. Retaliating with equal tariffs could ignite a full-blown trade war that disrupts supply chains, increases production costs, and dampens export demand.

Economic models from the Bank of Canada suggest that if the U.S. imposes 25% tariffs and Canada mirrors with its own 25% cuts on U.S. imports, GDP growth could drop by about 2.5 percentage points in the first year and 1.5 percentage points in the second, a cumulative decline of nearly 4% relative to a no-tariff scenario.

This significant slowdown would force Canadian companies to scale back hiring while consumers face even steeper price hikes as higher costs ripple through the economy.

History shows that trade wars rarely yield winners; they tend to lock economies into long-term instability and hardship. Instead of triggering a retaliatory spiral, the US and Canada should focus on smart, diplomatic negotiations to protect their industries and maintain open trade channels.

[Ajithdhev @ajithdhev]

A tit-for-tat strategy only escalates the pain for everyone involved. While it's easy to vent frustration, retaliatory tariffs or boycotts don't solve the underlying issues—they just hurt both Canadian and American consumers and businesses. In a trade war, no one really wins; higher import fees, weakened currencies, and reduced economic growth create a lose-lose scenario for all. Negotiating in good faith is the ideal solution.

[Ajithdhev @ajithdhev]

US 25% tariffs on Canadian goods could spike prices, slash jobs, and weaken the loonie. When steep tariffs are imposed, everyday costs rise as imported goods become more expensive—squeezing family budgets and putting small businesses under pressure.

But the real danger isn’t just higher prices—it’s how Canada might respond. Retaliating with equal tariffs could ignite a full-blown trade war that disrupts supply chains, increases production costs, and dampens export demand.

Economic models from the Bank of Canada suggest that if the U.S. imposes 25% tariffs and Canada mirrors with its own 25% cuts on U.S. imports, GDP growth could drop by about 2.5 percentage points in the first year and 1.5 percentage points in the second, a cumulative decline of nearly 4% relative to a no-tariff scenario.

This significant slowdown would force Canadian companies to scale back hiring while consumers face even steeper price hikes as higher costs ripple through the economy.

History shows that trade wars rarely yield winners; they tend to lock economies into long-term instability and hardship. Instead of triggering a retaliatory spiral, the US and Canada should focus on smart, diplomatic negotiations to protect their industries and maintain open trade channels.

[Ajithdhev @ajithdhev]

You're absolutely right. When tariffs are imposed, the currency can weaken, which then makes imports more expensive. That increase in import fees means everyday goods cost more, and small businesses already operating on tight margins feel the squeeze even harder. This cycle of higher prices and reduced purchasing power can ultimately hurt the broader economy.

[Ajithdhev @ajithdhev]

Over the weekend, I delved into DeepSeek-R1 and found it impressive. However, I'm bracing myself for what'll come up within the next few months - that'll be truly disruptive! Here's what I understand so far about how the model works (in a very simplified manner) :

DeepSeek R1 is an open-source AI model that matches the performance of OpenAI's o1 in reasoning tasks.

Chain of Thought (CoT) prompting encourages the model to explain its reasoning step by step, enhancing its problem-solving abilities.

Reinforcement Learning in DeepSeek R1 involves selecting the most effective problem-solving strategies by evaluating different approaches (Policy) and choosing the ones with the highest rewards.

Group Relative Policy Optimization (GRPO) is a technique used in DeepSeek R1 to refine its reasoning methods incrementally, ensuring stable and consistent improvements.

Distillation involves training a smaller "student" model to mimic a larger "teacher" model, resulting in an efficient AI that requires fewer resources while maintaining high performance.

DeepSeek R1's efficient design allows it to perform competitively with leading AI models, even with reduced computational requirements DeepSeek-R1 has truly democratized access to advanced AI.

Exciting times!

[Ajithdhev @ajithdhev]

@solodololoner14 I’m hopeful that OpenAI would release it by the end of March

[Ajithdhev @ajithdhev]

AI Agents are the future!

I built this in 2 hours on a Sunday evening using Claude because I was bored and I don’t even code full-time. Now think what would happen once we have the most hyped o3 model!

The future is equal parts exciting and terrifying.