chandureddyvari

You can’t compare these with regular VM of aws or gcp. VM are expected to boot up in milliseconds and can be stopped/killed in milliseconds. You are charged per second of usage. The sandboxes are ephemeral and meant for AI coding agents. Typical sandboxes run less than 30 mins session. The premium is for the flexibility it comes with.

For anyone curious about what LEAN is, like me, here’s the explanation: Lean Theorem Prover is a Microsoft project. You can find it here: https://www.microsoft.com/en-us/research/project/lean/

There were other HN posts suggesting BMAD, ccpm, conductor, etc. I considered giving it a try. They were quite comprehensive, to the point where I was exhausted reading all the documentation they’ve generated before coding - product requirements, epics, user stories/journeys, tasks, analysis, architecture, project plans.

The idea was to encapsulate the context for a subagent to work on in a single GitHub issue/document. I’m yet to see how the development/QA subagents will fare in real-world scenarios by relying on the context in the GitHub issue.

Like many others here, I believe subagents will starve for context. Claude Code Agent is context-rich, while claude subagents are context-poor.

Unsolicited advice: Why doesn’t open router provide hosting services for OSS models that guarantee non-quantised versions of the LLMs? Would be a win-win for everyone.

I use Roo code with orchestrator(Boomerang) mode which pretty much has similar workflow. The orchestrator calls the architect to design the specs, and after iterating and agreeing on the approach, it is handed over to Code mode to execute the tasks. Google Gemini 2.5 pro is pretty good at orchestration due to its 1M context and I use claude sonnet 4 for code mode.

What else does Kiro do differently?

Edit: The hooks feature looks nifty. How is the memory management handled? Any codebase indexing etc? Support to add external MCP servers like context7 etc?

Getting rate limit error - This request would exceed the rate limit for your organization

You should use something like openrouter or portkey or similar for managing fallbacks

I’m exploring two different applications of AI for education and skill-building:

1. Open-Source AI Curriculum Generator(OSS MathAcademy alternative for other subjects) Think MathAcademy meets GitHub: an AI system that generates complete computer science curricula with prerequisites, interactive lessons, quizzes, and progression paths. The twist: everything is human-reviewed and open-sourced for community auditing. Starting with an undergrad CS foundation, then branching into specializations (web dev, mobile, backend, AI, systems programming).

The goal is serving self-learners who want structured, rigorous CS education outside traditional institutions. AI handles the heavy lifting of curriculum design and personalization, while human experts ensure quality and accuracy.

2. Computational Astrology as an AI Agent Testbed For learning production-grade AI agents, I’m building a system that handles Indian astrology calculations. Despite the domain’s questionable validity, it’s surprisingly well-suited for AI: complex rule systems, computational algorithms from classical texts, and intricate overlapping interpretations - perfect for testing RAG + MCP tool architectures.

It’s purely a technical exercise to understand agent orchestration, knowledge retrieval, and multi-step reasoning in a domain with well-defined (if arcane) computational rules.

- Has anyone tackled AI generated curricula? What are the gotchas? - Interest in either as open-source projects?

Thanks just discovered https://ai.pydantic.dev/graph/

from the above link, which seems to use FSMs instead of DAGs.

Neat idea! It would be helpful to have LLMs ranked from best to worst for a given GPU. Few other improvements I can think of:

- Use natural language for telling offloading requirements.

- Just year of the LLM launch of HF url can help if it’s an outdated LLM or a cutting edge LLM.

- VLMs/Embedding models are missing?

Wow. Hate is unwarranted. Tirumala Anna Prasadam is pretty good. quoting from press release “By the normal standards in the present kitchen 12 huge vessels are steamed to produce rice at the rate of five persons for one kg rice. “But we are geared to cook rice for one to two lakh persons during the Srivari Brahmotsavam at the rate of 400 bags of rice per day”

https://news.tirumala.org/nitya-annadam-free-meals-for-all-2...

more stats here

https://www.tirumala.org/SRIVENKATESWARAANNAPRASADAMTRUST.as...

Anna prasadam is pretty common thing in Hinduism. Read about Akshayapatra foundation which provides free mid day meals to 1000’s of children every single day!

Steve Jobs would walk 7 miles every Sunday to have lunch at Hare Krishna Temple

https://x.com/iskconinc/status/1305841447711563776?s=46

I'm working on helping my wife get her print-on-demand Shopify store off the ground. She designs the products herself, but ran into challenges with SEO. So, I built a custom app that connects to her Shopify store via API, using GPT-4o-mini to handle SEO optimizations—things like generating descriptions, titles, alt text, SEO-friendly URLs, and more, all based on the product images.

There were also issues syncing with Google Merchant Center (missing colors, categories, etc.), so I tweaked the app to auto-fill these fields using GPT-4o, making it compliant with Google and Pinterest requirements.

I’m completely new to SEO and just tried following best practices to fix things as they came up. Now I’m learning that SEO keywords change constantly, so I’m thinking of integrating a keyword provider to dynamically enhance our product descriptions.

I never realized running an e-commerce store (especially print-on-demand) would involve so much operational work on the marketing front. I’d appreciate any advice on what to tackle next, especially since my goal is to avoid “subscription hell” with multiple Shopify apps. My wife is also just starting with ads and campaigns, diving into tutorials to learn the ropes.

Any guidance is welcome!

I'll try to slightly dive into Advaita philosophy here. Both the Upanishads and the Yoga Vasistha affirm that the distinction between the seer and the Self is an illusion created by the mind. When this illusion is dispelled, the oneness of the Self is realized. The intellect, which seems to function as a separate tool, is ultimately part of the same illusion. True knowledge is realizing that there is no separation—everything is the Self.

Quoting one from Katha Upanishad 1.3.10:

"Indriyebhyaḥ parā hy artha, arthebhyas ca param manah, manasas tu parā buddhir buddher ātmā mahān parah"

Translation: "Beyond the senses are the objects, beyond the objects is the mind, beyond the mind is the intellect, beyond the intellect is the Self."

This quote emphasizes that the intellect is still a part of the illusion. Beyond even the intellect lies the Self, which is one and undivided.

There’s a relevant quote from Swami Vivekananda on this

From Karma Yoga, Chapter: I, Karma in its effect on character—

What we say a man “knows”, should, in strict psychological language, be what he “discovers” or “unveils”; what a man “learns” is really what he “discovers”, by taking the cover off his own soul, which is a mine of infinite knowledge.

We say Newton discovered gravitation. Was it sitting anywhere in a corner waiting for him? It was in his own mind; the time came and he found it out. All knowledge that the world has ever received comes from the mind; the infinite library of the universe is in your own mind. The external world is simply the suggestion, the occasion, which sets you to study your own mind, but the object of your study is always your own mind. The falling of an apple gave the suggestion to Newton, and he studied his own mind. He rearranged all the previous links of thought in his mind and discovered a new link among them, which we call the law of gravitation. It was not in the apple nor in anything in the centre of the earth.

Whenever I read about Ramanujan having divine revealing formulas in his dreams, I remember Swami Vivekananda’s quoute on consciousness and mind.

edit: found another relevant quote from Upanishads on tapping the infinite knowledge:

Mundaka Upanishad 2.2.9:

“Eṣa sarveṣu bhūteṣu gūḍhātmanā prakāśate, dṛśyate tvagryayā buddhyā sūkṣmayā sūkṣmadarśibhiḥ”

Translation: “The Self hidden in all beings does not shine forth, but it is seen by subtle seers through their one-pointed and subtle intellect.”

Explanation: The ultimate knowledge or truth is hidden within all beings and is revealed through subtle inner perception. The idea is knowledge is latent within the mind, and it is discovered, not externally found.

This article about dealing with recurrence in applications was an eye-opener for me. Definitely recommend giving it a read. https://github.com/bmoeskau/Extensible/blob/master/recurrenc...

Also fungibility! I’m equally concerned about the specific “purpose“ and “time duration”, the Deputy T Governor gave an example to the press., quoting from the article:

“Let’s say a school has given money to a student who won a prize to buy books in a particular bookshop. So the student can only use this money for that purpose.

But as soon as the book is purchased and the currency goes to the bookshop owner, it becomes fungible again.

So it’s only for that period the fungibility of money becomes limited.

If the student chooses not to spend the money, it goes back to the school and it becomes fungible again.

So programmability does not militate against fungibility but rather, just puts it on hold.”

The CBDC(Central Bank Digital Currency) Retail (CBDC-R) pilot currently enables Person to Person (P2P) and Person to Merchant (P2M) transactions. It is now proposed to enable additional functionalities of programmability and offline capability in CBDC retail payments. Programmability will facilitate transactions for specific/targeted purposes, while offline functionality will enable these transactions in areas with poor or limited internet connectivity.

Sorry if this is a dumb question. Can someone explain why it’s called 8x7B(56B) but it has only 46.7B params? and it uses 12.9B params per token generation but there are 2 experts(2x7B) chosen by a 2B model? I’m finding it difficult to wrap my head around this.

I had lot of fun chatting with Pi. After some poking around to give it’s “system prompt”(long when it came back and prompt injection was a cool thing)., it said it was using some conversational frameworks like Grice’s principle etc. I tried to recreate one in GPT store. I call it Tara. Try it here- https://chat.openai.com/g/g-mI1QatRrc-tara.

Interesting. I thought anything >1million would need a vector db to scale on production. What was your machine config for running faiss? Also did you plan for redundancy or was it just faiss as a service VM?

Yeah what you mentioned might be true. Currently our understanding on how LLMs really work behind the screens is limited. For example, there was a recent research[1] where LLM's accuracy is better if the context is added at the beginning when compared to the end of the prompt. So it's mostly by trial & error to figure out what works out best for you. You can use FAISS or similar to have the embeddings in-memory instead of a full fledged vector DB. But pg vector is convenient plugin if you already have postgres instance running

[1]- https://towardsdatascience.com/in-context-learning-approache...

The context length is limited, for gpt-3.5 it's 4k tokens, there are other offerings which offer upto 100k(claude). 100k tokens is ~1 book., but it priced steeply for each call. It's often wiser, cheaper to Retrieve the context from your text & Augment your query to the LLM to Generate more contextual answers. That's the reason for the name Retrieval Augmented Generation (RAG). For Retrieving - you'd need a vector database (for similarity comparison you can use semantic or vector embedding based similarity search)

slightly tangential, but where do people get awesome landing pages like linear(https://fig.io/. has similar landing page) etc. Do they build them in-house or buy templates somewhere? Many of the recently launched YC companies have awesome landing pages. eg. https://automorphic.ai/,

yeah. my bad. Turning off relay, did show the right message. Thanks for that.

It sounds great. But their banner is showing that my ip address is from Mumbai, whereas I’m actually in Bengaluru, India. That’s not really re-assuring. Maybe it’s just apple relay on my device that’s obfuscating my details. edit: my bad, hit me bit late that it’s the intended behaviour.

Interesting project. How does it fare on scaling? We were evaluating vector dbs and since we were into b2b saas, we are keen on the sharding, scaling and multi tenancy features. currently we are more inclined towards milvus. https://zilliz.com/comparison/qdrant-vs-milvus

FrugalGPT is a simple yet flexible instantiation of LLM cascade which learns which combinations of LLMs to use for different queries in order to reduce cost and improve accuracy. Experiments show that FrugalGPT can match the performance of the best individual LLM (e.g. GPT-4) with up to 98% cost reduction or improve the accuracy over GPT-4 by 4% with the same cost

I’ve been using Screen zen for quite some time now. I found it to be incredibly useful. Curious to know how clearspace is different from screenzen, given the latter is free on app store. https://apps.apple.com/in/app/screenzen-screen-time-control/...

Working for me on Firefox Mac.

Wow. Nice! Didn't know this exists. Planning to get one. Do you have any recommendations?

I really like your learning series. What you’ve done for understanding conversational memory at https://www.pinecone.io/learn/langchain-conversational-memor... was truly helpful. Thanks!