Shashank Srikanth

ML Platform @ Netflix. OSS Contributor @ Metaflow. Interests: ML Infra, Computer Vision, & Investing.
Personal Blog:

[Shashank Srikanth @s_shawshank]

@jacobwillemsma People like to do things that they think and feel are cool. Nothing wrong with it.

[Shashank Srikanth @s_shawshank]

@srush_nlp So this is the future of science 😂

[Shashank Srikanth @s_shawshank]

@GergelyOrosz But if Muse Spark is available via API, then you can just use it with your own harness, like PI, Hermes, Cursor, and more. You don't always need to build the harness/developer infra to get adoption.

[Shashank Srikanth @s_shawshank]

@scaling01 @zephyr_z9 Do you really need Mythos level models for all your coding tasks? GPT 5.5 level models are more than good enough for a lot of the tasks.

[Shashank Srikanth @s_shawshank]

@jmduke Slop everywhere. At least put some effort to not make it look like slop.

[Shashank Srikanth @s_shawshank]

@ariG23498 Super cool - loved the first part, since most tutorials jump straight into nvidia-insights and CUDA/kernel-level optimizations without covering the basics.

[Shashank Srikanth @s_shawshank]

@simonw @PetuniaByte With all due respect, why respond to a clearly AI generated post. It's not worth your time.

[Shashank Srikanth @s_shawshank]

@0xlelouch_ This post itself reeks/sounds like AI slop with all due respect.

[Shashank Srikanth @s_shawshank]

@blondesnmoney Even if an AI could help with a particular query, why would I ever call it? The chat interface is simply superior and faster - you don't have to remember the agent's answers, the entire output will be in front of you with clickable links/filled forms, etc.

[Shashank Srikanth @s_shawshank]

@blondesnmoney If I wanted to know the prices/service options, I would just look online. The only reason to call customer service is to seek help on OOD queries/requests that require explicit human approval. If AI can help with that, great! Otherwise, I would rather be routed to a human ASAP.

[Shashank Srikanth @s_shawshank]

@abhi1thakur PyCharm is still ahead in code navigation and debugger options. But yeah, with the rise of Cursor et al, I don’t think the future is bright for IntelliJ/Pycharm.

[Shashank Srikanth @s_shawshank]

@emilyinvc Ragebait?? 5K for an angel check?

[OpenAI @OpenAI]

In the coming weeks, we plan to start testing ads in ChatGPT free and Go tiers.

We’re sharing our principles early on how we’ll approach ads–guided by putting user trust and transparency first as we work to make AI accessible to everyone.

What matters most:
- Responses in ChatGPT will not be influenced by ads.

- Ads are always separate and clearly labeled.

- Your conversations are private from advertisers.

- Plus, Pro, Business, and Enterprise tiers will not have ads.

[Shashank Srikanth @s_shawshank]

@MikeIsaac Apple wants a white-label product (see Gemini deal), and OpenAI is probably opposed to it, as they want to maintain a direct relationship with their customers, since they are both an AI research lab and a consumer tech company.

[Shashank Srikanth @s_shawshank]

@Ajay_Bagga https://t.co/5jWKPA4N05

This @SemiAnalysis_ post provides a nice overview of how consumer LLM-based chat apps can tackle monetization. There's a lot more complexity involved, like auctions, recommendations, privacy, but the direction is clear.

[Shashank Srikanth @s_shawshank]

@Ajay_Bagga Well, LLM search will expand the market by making some queries more valuable vs search: “best credit card” → “best card for my spending pattern”

Also, inference is becoming cheaper &
Routing models (GPT5) can reduce costs, e.g.
Simple Query -> GPT5-nano
Complex Query: GPT5-Pro

[Shashank Srikanth @s_shawshank]

@Duderichy I believe that python dicts were already ordered in 3.6 onwards (implementation detail) but only got added to the spec in 3.7. Also, ordered dicts allow you to reorder elements unlike the normal one so there’s still reasons to continue using them.

[Shashank Srikanth @s_shawshank]

@nearcyan LMArena was literally a streamlit app as of a year ago.

[Shashank Srikanth @s_shawshank]

@ShashwatGoel7 @deepigoyal I wouldn’t trust tools like these to be accurate most of the times, but in this case, it’s quite evident that it’s AI generated.

[Shashank Srikanth @s_shawshank]

@deepigoyal If you are going to generate AI slop, at least use Gemini / Claude, so that it’s not so obvious that this was AI generated.