Shodhak

Sharing my views on Tech and investments

[Shodhak @ShodhakNil]

@elonmusk You will get government pressure to remove X fact check feature. Please don’t remove it, fact check feature and @grok is what keeps the X a honest place to share information. Hope we can count on you

[Shodhak @ShodhakNil]

When deciding whether to do something, compare it with the best opportunities you have

[Shodhak @ShodhakNil]

People getting laid off from these big companies think AI ate their job because till now they were a very small part of a big project, they never got the full idea what really is happening. I observe people working at an MNC and startups have a completely different mentality. Startups move fast, you take more ownership that’s why you know all loop holes where AI can be used and cannot, software engineers at startups are till working on the logic then the code, AI can write better code if the prompt is optimised, a custom harness is what every engineer should aim for, this will give different output even if everyone is using the same model.

[Shodhak @ShodhakNil]

Everyone is focusing on building models, deploying them but they realise that the basic transformer architecture will never be cost efficient. We need to change the architecture, to make models energy and cost efficient. They will only solve the real human problems if they run continuously in loop finding pattern missed by humans and to do so change in architecture is required(Chinese companies did it with limited hardware US companies can do it too, if they try), we cannot just add up additional gpus, there such a high latency issue between gpu-to-gpu communication that we can’t overcome soon. Faster loop continuous and 0 drop is the ideal solution.

[Shodhak @ShodhakNil]

@SejalSud Depends, main Mumbai around 5-6L post tax minimum

[Shodhak @ShodhakNil]

Soon we will see a lot of startups open sourcing their SLM, enterprise will adopt around these startups because it’s way too cheap and efficient to run and more accurate compared to a LLM with less active parameters. A replacement around MCP will be open-sourced, probably in start of 2028

[Shodhak @ShodhakNil]

Did a random experiment on sleep. For context I usually have my dinner between 16:30 to 17:30. But yesterday I had a heavy meal around 20:00, my sleep score was decreased and guess what my heartrate was above daily avg during sleep. This shows how late night meals can affect your sleep, heart and eventually the overall recovery of your body. Sharing my Garmin metrics ss. The avg heartbeat changed from 47 to 67bpm.

[Shodhak @ShodhakNil]

@mfranz_on For 128GB both cost the same

[Shodhak @ShodhakNil]

@samongaro_ Flirting in real life

[Shodhak @ShodhakNil]

@RonnieScrewvala @origin_autonomy The hardware development cycle is very long, even if we start now it will take atleast 7-8 yrs (post RnD) to build something meaningful that world would use.

[Shodhak @ShodhakNil]

@digitalix After @elonmusk @digitalix is the richest person 🫠

[Shodhak @ShodhakNil]

Every government should respect their youth the young population, know why? Because this population is going to vote for next 50 yrs and they will remember who did what, this generation believes in asking questions and you must answer them with full honesty. This youth wants the elected representatives to be honest and neutral towards all groups. They cannot be brain washed because they question why.

[Shodhak @ShodhakNil]

@Taniyatweets_ I want mac hardware with linux so i can run my x11. I hate macOS and wayland

[Shodhak @ShodhakNil]

We should think of LLM code as a compiler that translates programming language into natural language, just an additional layer of how tech evolves.

[Shodhak @ShodhakNil]

@konusisco There’s so much traffic in the sky

[Shodhak @ShodhakNil]

@anah_sahh They want US to collect their data

[Shodhak @ShodhakNil]

😂 People won’t be able to architect the complete project. AI is bad if we don’t provide the right technical input. Cost to generate software was always low, architecting a product was high and will be high with AI as well, people who knows how to leverage and ask AI to build will be in demand

[Shodhak @ShodhakNil]

I’m a heavy claude code user, why should I use another instance for my agent if Im in development stage. For prod I will have a secure runtime but for dev this doesn’t make any sense. Also the cost of infra will be really high in future, but I personally don’t see any need for this. I maybe wrong

[Shodhak @ShodhakNil]

Opensource software is changing, companies till now contributed to major projects which they had in prod at scale. They promoted opensource because of lack of resources, they cannot focus on just maintaining, but now with AI why would a company promote open source software instead they can build a more coupled software with their system and maintain using AI agents. The quality will be above average, and the same people taking calls in community meets, will internally orchestrate the project. Yes Opensource won’t be same, we need something more

[Shodhak @ShodhakNil]

Change is everything, We humans have a great ability to adapt. With this AI wave we need to change how we think and work, next wave will robots and we cannot deny that, no government will oppose this. The world revolves around parameters, current is money next will be data. We still don’t have enough data for LLMs to predict our lives, and we beed that for next wave, with second wave the parameter money will be gone, everyone will have enough resources to survive but parameter will change so the governing law and order. Next 15 yrs is something that humans have never imagined. First 5yrs from now that’s 2032 will have enough stability and optimal LLM security, they will be everywhere trust me everywhere. 2033 they will hit a threshold and robots will take over, labour doing repetitive work will be replaced, advance machinery will be built by them, and everything will be controlled by LLMs, humans will be the inital and final loop. The final decision will always be human.