Hacker News: manojbajaj95

New comment by manojbajaj95 in "Astra for Coding: Why Are We Doing This Again?"

manojbajaj95 — Fri, 11 Sep 2026 08:34:04 +0000

I've been extremely frustrated with any large new work that i do with agents. Then plan multi step, multi hour work with extremely large code changes running for 30+ hours. In the end what you get is sometime completely useless code because it made an assumption that wasn't true at all. In the end, i end up wasting hours.

Astra for Coding: Why Are We Doing This Again?

manojbajaj95 — Fri, 11 Sep 2026 06:23:06 +0000

Article URL: https://lucumr.pocoo.org/2026/9/7/astra-why/

Comments URL: https://news.ycombinator.com/item?id=49654229

Points: 361

# Comments: 263

We Designed Muse

manojbajaj95 — Wed, 09 Sep 2026 06:27:07 +0000

Article URL: https://introducing.muse.ai

Comments URL: https://news.ycombinator.com/item?id=49621990

Points: 1

# Comments: 0

Qwen/Qwen3.8-Flash-Next · Hugging Face

manojbajaj95 — Thu, 27 Aug 2026 04:50:14 +0000

Article URL: https://huggingface.co/Qwen/Qwen3.8-Flash-Next

Comments URL: https://news.ycombinator.com/item?id=49459873

Points: 3

# Comments: 0

Google introduces Pixel 11, Pixel 11 Pro and Pixel 11 Pro XL

manojbajaj95 — Wed, 12 Aug 2026 16:19:18 +0000

Article URL: https://blog.google/products-and-platforms/devices/pixel/google-pixel-11-pro-xl/

Comments URL: https://news.ycombinator.com/item?id=49274837

Points: 13

# Comments: 3

New comment by manojbajaj95 in "Don't start a startup with the goal of getting acquired"

manojbajaj95 — Tue, 04 Aug 2026 09:14:52 +0000

Short sweet and to the point.

New comment by manojbajaj95 in "Harness Engineering for Self-Improvement"

manojbajaj95 — Tue, 04 Aug 2026 09:11:50 +0000

I've been working with self improvement harness a little bit and one thing i've come to conclusion is harness task fit. The learning can be significantly improved if we understand the behaviour of task and how it should be learned. I'm pretty sure a general solution will definitely exist which will do fine, but we are yet to see one.

New comment by manojbajaj95 in "LLMs reward expertise"

manojbajaj95 — Tue, 04 Aug 2026 09:07:15 +0000

I agree with the premise that LLMs reward experstise, but people without expertise can very eaily learn to prompt correctly and get to a result that is very good. I remember somebody proved a mathematical conjecture by just asking 'keep going' in plain english without a mathematics background.

New comment by manojbajaj95 in "Clanker: A Word for the Machine"

manojbajaj95 — Wed, 27 May 2026 09:42:26 +0000

Strong agree with the point that "agents" shouldn't be called "agents", as agency lies with human. That being said, I don't like the word clanker either.

New comment by manojbajaj95 in "Show HN: Agent.email – sign up via curl, claim with a human OTP"

manojbajaj95 — Fri, 22 May 2026 07:51:27 +0000

What are some of the use cases where AgentMail makes sense?

New comment by manojbajaj95 in "Show HN: InsForge – Open-source Heroku for coding agents"

manojbajaj95 — Mon, 18 May 2026 21:32:58 +0000

Congrats on the launch. Currently all my hobby projects go like procure auth from clerk, database from neon, vercel/cloudflare for deployment, etc etc. This could actually solve that problem. Is there a local simulator that can be used for testing and then replicated easily in prod?

New comment by manojbajaj95 in "Show HN: AgentPort – Open-source Security Gateway For Agents"

manojbajaj95 — Thu, 30 Apr 2026 07:29:57 +0000

How does this differ from multiple MCP gateways that are available? And how would this work with our custom tools?

New comment by manojbajaj95 in "Authsome – open-source local auth proxy for AI agents"

manojbajaj95 — Wed, 29 Apr 2026 05:57:44 +0000

Haven't used clawvisor, but what I understand is that its trying to solve the same problem but very diffefrently. Authsome doesn't bundle any tooling and works as an MITM proxy with credential injection. This means you can work with your own tools, cli's or MCPs.

New comment by manojbajaj95 in "Show HN: Agent Vault – Open-source credential proxy and vault for agents"

manojbajaj95 — Mon, 27 Apr 2026 13:06:48 +0000

I'm building https://github.com/manojbajaj95/authsome to solve similar problem. Would love to know your opinion.

New comment by manojbajaj95 in "Show HN: Agent Vault – Open-source credential proxy and vault for agents"

manojbajaj95 — Fri, 24 Apr 2026 08:23:54 +0000

Try out, https://github.com/manojbajaj95/authsome . I'm trying to work with LLM secrets too. Still alpha, but thats the direction i'm taking

New comment by manojbajaj95 in "Show HN: Agent Vault – Open-source credential proxy and vault for agents"

manojbajaj95 — Fri, 24 Apr 2026 07:54:05 +0000

Been tinkering with something of my own at https://github.com/manojbajaj95/authsome. Core goal was to do credential management, from an ease point of view and not security.

New comment by manojbajaj95 in "Yann LeCun raises $1B to build AI that understands the physical world"

manojbajaj95 — Tue, 10 Mar 2026 19:57:52 +0000

I attended a talk from Yann LeCun, and he always had a strong opinion about auto-regressive models. Its nice to see someone not just chasing hype and doing more research.

New comment by manojbajaj95 in "Show HN: MCP-skill – generate Python skills from MCP servers"

manojbajaj95 — Fri, 06 Mar 2026 13:29:13 +0000

Happy to answer questions about how the code generation works or how this integrates with Claude Code.

Show HN: MCP-skill – generate Python skills from MCP servers

manojbajaj95 — Fri, 06 Mar 2026 13:20:15 +0000

I've been experimenting with MCP-based agents for some of my daily workflows (checking assigned issues, gathering context, drafting plans, etc.).

With most agent frameworks, tool calls run in a loop like this:

agent → tool → agent → tool → agent → tool

Each step goes back through the model. For tools returning large payloads, this adds latency and burns tokens.

So I built a small tool that converts an MCP server into a Python skill.

It introspects the server and generates a typed Python class where each tool becomes an async method.

Example:

    issue = await github.get_issue(issue_id)
    docs = await search.web_search(issue.title)

The agent can process results locally in code and only return the final output.

This also enables progressive disclosure. Tools can return large responses, but the agent can filter or aggregate data in code instead of sending everything back into the model context.

The generated skills include a `SKILL.md` file and can be reused later, so useful workflows can be saved and run again as scripts.

Still a work in progress, but already useful in my own setup.

Would love feedback from others experimenting with MCP or similar patterns.

Repo: https://github.com/manojbajaj95/mcp-skill

Comments URL: https://news.ycombinator.com/item?id=47274589

Points: 2

# Comments: 1

New comment by manojbajaj95 in "Nobody gets promoted for simplicity"

manojbajaj95 — Wed, 04 Mar 2026 06:59:56 +0000

Early in my carrier, i saw code written by people both junior and senior. Every single time i saw a great code, it made me feel like its so simple. heck, i could have written it. It was completely opposite for junior-mid folks.