chandureddyvari

Took me a bit to understand what Wire does, but then it clicked. we’ve built something similar at a smaller scale inside R2, though right now it’s only for .md files.

I can see this becoming much more useful once the docs get heavier: large PDFs, XLSX files, images, etc. At that point you probably need embeddings, reranking etc. But I think agents are smart enough to write scripts to retrieve what they want if we run them on a sandbox(which we are trying to do currently). bookmarking this for now.

Good write-up though!

I’m currently cobbling sub agents with hooks, workflows looks very promising for doing things more predictably.

Is this equivalent of DAGs for sub agents inside claude code? Can i pause and resume/retry workflows? How stateful are they?

Really appreciate it someone claude code can throw more light on above. I’m trying to see if I can get langgraph equivalent DAGs here.

Wasn't Mythos a step change improvement?

What’s your favourite harness? Is there any benchmarks for harness like LLMs have for swe verified?

I was to talking to a YC founder, his biggest fear is waking up to a new Claude launch making his startup obsolete the next morning.

Similar sentiment shared with other startup founders- check on x about all VCs talking about moats against big labs.

or the word 'canonical'

some of them are non existent today. Check the parent thread - some good recommendations(for 2023) on both functional websites and pretty websites. At that time if I recall linear landing page was all rage, and there were many copycats.

For a long time I wondered how SV startups got such pretty landing pages (here’s a comment I left 2 years back: https://news.ycombinator.com/item?id=37421273). I wanted one for my side projects but couldn’t afford an agency, and the templates online were boring. Creating the page was only half the problem. I also needed somewhere to collect emails for the waitlist.

After AI happened, I built an app (promptfunnels) to scratch my own itch and generate funnels (fancy name for landing pages with a purpose).

Then came the harder part: marketing it. Coming from a tech background, I knew nothing about marketing, so I started reading and came across the $100M Leads book. I realized codifying those principles together with funnels and marketing automation had a real market. My family, friends, and acquaintances became the first customers. A friend joined me as cofounder and we both quit our jobs to do this full time.

As we talked to other startup founders, they kept describing a tangential problem they called GTM. At the core it was the same thing we were solving: marketing for non-marketers. So we pivoted to RevMozi(https://revmozi.com/), which helps non-marketers do both inbound and outbound GTM.

We’re dogfooding the product and coming out of beta next month.

Wish us luck.

I had good success with hooks in claude code. Personally I feel this problem was common with humans as well. We added tools like husky for git commits, for our peers to push code which was linted, type checked etc.

I feel hooks are integral part of your code harness, that’s only deterministic way to control coding agents.

For me it’s mostly useful in day-to-day coding, not “build an entire app and walk away” coding.

TDD was never really my natural style, but LLMs are great at generating the obvious test cases quickly. That lets me spend more of my attention on the edge cases, the invariants, and the parts that actually need judgment.

Frontend is another area where they help a lot. It’s not my strongest side, so pairing an LLM with shadcn/ui gets me to a decent, responsive UI much faster than I would on my own. Same with deployment and infra glue work across Cloudflare, AWS, Hetzner, and similar platforms.

I’m basically a generalist with stronger instincts in backend work, data modeling, and system design. So the value for me is that I can lean into those strengths and use LLMs to cover more ground in the areas where I’m weaker.

That said, I do think this only works if you’re using them as leverage, not as a substitute for taste or judgment.

Claude has gotten noticeably worse for me too. It goes into long exploration loops for 5+ minutes even when I point it to the exact files to inspect. Then 30 minutes later I hit session limits. Three sessions like that in a day, and suddenly 25% of the weekly limit is gone.

I ended up buying the $100 Codex plan. So far it has been much more generous with usage and more accurate than Claude for the kind of work I do.

That said, Codex has its own issues. Its personality can be a bit off-putting for my taste. I had to add extra instructions in Agents.md just to make it less snarky. I was annoyed enough that I explicitly told it not to use the word “canonical.”

On UI/UX taste, I still think current Codex is behind the Jan/Feb era of Claude Code. Claude used to have much better finesse there. But for backend logic, hard debugging, and complex problem-solving, Codex has been clearly better for me. These days I use Impeccable Skillset inside Codex to compensate for the weaker UI taste, but it still does not quite match the polish and instinct Claude Code used to have.

I used to be a huge Claude Code advocate. At this point, I cannot recommend it in good conscience.

My advice now is simple: try the $20 plans for Codex and Cursor, and see which one matches your workflow and vibes best

Maybe I’m in the minority here, but while directories and similar channels are useful, I felt like I was just shooting darts in the dark without understanding sales and marketing from first principles and hoping something would stick.

I had three side projects and kept struggling to get any real traction or traffic without becoming spammy across the internet. So I decided to approach it the same way I approach learning anything new: through books, courses, and solid foundational material.

HN had a few excellent suggestions. One of them was Founding Sales. Another, which I came across through a friend’s recommendation, was Alex Hormozi’s series. He seems to have something of a cult following, which made me a bit skeptical at first, so I decided to just read the first 100 pages before forming an opinion.

I ended up finding it genuinely useful, especially for understanding the psychology and mindset needed to sell something. I now highly recommend his book $100M Leads to technical friends who are trying to figure out how to sell what they’ve built.

I’m still learning, if you’ve any good recommendations, please drop them below

Agreed, had the same experience. Codex feels lazy - I have to explicitly tell it to research existing code before it stops giving hand-wavy answers. Doc lookup is particularly bad; I even gave it access to a Context7 MCP server for documentation and it barely made a difference. The personality also feels off-putting, even after tweaking the experimental flag settings to make it friendlier.

For people suggesting it’s a skill issue: I’ve been using Claude Code for the past 6 months and I genuinely want to make Codex work - it was highly recommended by peers and friends. I’ve tried different model settings, explicitly instructed it to plan first and only execute after my approval, tested it on both Python and TypeScript backend codebases. Results are consistently underwhelming compared to Claude Code.

Claude Code just works for me out of the box. My default workflow is plan mode - a few iterations to nail the approach, then Claude one-shots the implementation after I approve. Haven’t been able to replicate anything close to that with Codex

Came across official anthropic repo on gh actions very relevant to what you mentioned. Your idea on scheduled doc updation using llm is brilliant, I’m stealing this idea. https://github.com/anthropics/claude-code-action

3b1b yt channel calculus & LA

https://explained.ai/matrix-calculus/

khan academy - Multivariable Calculus course by Grant Sanderson(3b1b fame)

Any good recommendations you got for learning kubernetes for busy people?

Is there a comprehensive leaderboard like ClickBench but for vector DBs? Something that measures both the qualitative (precision/recall) and quantitative aspects (query perf at 95th/99th percentile, QPS at load, compression ratios, etc.)?

ANN-Benchmark exists but it’s algorithm-focused rather than full-stack database testing, so it doesn’t capture real-world ops like concurrent writes, filtering, or resource management under load.

Would be great to see something more comprehensive and vendor-neutral emerge, especially testing things like: tail latencies under concurrent load, index build times vs quality tradeoffs, memory/disk usage, and behavior during failures/recovery

I have watched 3 animes- Solo levelling, Full metal alchemist and Attack on Titans. I liked them.

Edit: I realised i watched Full metal alchemist: Brotherhood didn’t know two versions existed

Pomelli is a new AI experiment from Google Labs that helps small-to-medium-sized businesses easily generate scalable, on-brand social media campaigns.

1. If is were possible for an ordinary mortal to impose arbitrary curses on the god of death and justice, the world would quickly descend into utter chaos.

Mandavya is not just any mortal; he is an enlightened sage. In Hinduism, enlightened beings are considered superior to gods. There’s another story about Sage Markandeya (one of the nine immortals, the Chiranjeevis) who caused the death of Yama, the God of Death. In Hindu cosmology, all the gods hold honorary responsibilities, and nothing is permanent - not even the position of Brahma, the Creator

2. If children are completely free from accountability, adults will form them into an army and convince them to commit crimes on their behalf, leading to an intolerable situation. This may already be a standard way of doing business in some parts of the world

I believe he introduced a juvenile law, which involves reduced sentences or milder punishments rather than granting complete immunity from consequences.

Slightly tangential but this was a learning moment for me.

This reminds me of a story where Sage Mandavya established the first juvenile law in Hindu mythology.

<story starts>

Long ago, there lived a great sage named Mandavya who had taken a vow of silence and spent his days in deep meditation. One day, while he sat motionless beneath a tree with his arms raised in penance, a group of thieves being pursued by the king’s soldiers fled into his hermitage. They hid their stolen loot near the sage and escaped through the other side. When the king’s soldiers arrived, they found the stolen goods but the sage—deep in meditation and bound by his vow of silence—neither confirmed nor denied their presence. The soldiers arrested him and brought him before the king, accusing him of harboring criminals.

Despite his spiritual stature, the king ordered a severe punishment: Mandavya was to be impaled on a stake (shula)—a horrific execution where a wooden spike was driven through the body. However, due to his immense yogic powers and detachment from the physical world, the sage did not die. He remained alive on the stake, enduring the agony with superhuman patience. Eventually, other sages intervened, the king realized his grave error, and Mandavya was freed. But the damage was done. When the sage finally left his mortal body, he went directly to Yamaloka—the realm of Yama, the god of death and justice—to demand an explanation.

“Why did I have to suffer such a gruesome fate?” Sage Mandavya asked Lord Yama. “What terrible sin did I commit to deserve impalement?” Yama consulted his records and replied, “When you were a child, you caught a dragonfly and pierced it with a needle through its body, watching it suffer for your amusement. That act of cruelty resulted in your punishment - you experienced the same suffering you inflicted on that innocent creature.”

Sage Mandavya was furious. “That was when I was a child!” he protested. “I was too young to understand the difference between right and wrong, between sin and virtue. How can you punish an ignorant child with the same severity as a knowing adult?”

Yama tried to explain that karma operates impartially, but Mandavya would not accept this. In his righteous anger, the sage cursed Yama himself: “For this unjust judgment, you shall be born as a human on Earth and experience mortality yourself!” This curse led to Yama being born as Vidura, the wise and virtuous counselor in the Mahabharata - a human who, despite his wisdom and righteousness, had to endure the limitations and sufferings of mortal life.

But Mandavya didn’t stop there. Using his spiritual authority, he proclaimed a new divine law: “No sin committed by a child below the age of fourteen shall count toward their karmic debt equivalent to that of an adult. Children who do not yet understand dharma and adharma shall not be punished for their ignorant actions.” This became the first “juvenile law” in Hindu mythology—a recognition that children, in their innocence and ignorance, deserve compassion and correction rather than severe punishment.

<story ends>

When I was a child, I too wanted to catch a dragonfly and tie a thread to it so it would fly around like a little pet. But my mother stopped me. She told me this very story of Sage Mandavya, and it scared me for life. I never forgot it, and I never tried to catch and bind a dragonfly again.

Nice, so the open source equivalent now exists. Meta basically commoditized Tinker's($12B valuation) value prop by giving away the infra (Monarch) and the RL framework (TorchForge). Will be interesting to see how a managed service competes with free + open source at this layer.

Interesting - this seems to target a different layer than services like Tinker (https://thinkingmachines.ai/blog/announcing-tinker/). Monarch provides the infrastructure primitives while Tinker is a managed finetuning service. Could someone build something like Tinker on top of Monarch?

In my case, it is ignorance. I am not familiar with how to wield firecracker VMs and manage their lifecycle without putting a hole in my pocket. These sandbox services(e2b, Daytona, Vercel, etc.) package them in an intuitive SDK for me to consume in my application. Since the sandboxing is not the main differentiator for me, I am okay to leverage the external providers to fill in for me. That said, I will be grateful if you can point me to right resources on how to do this myself :)

Can someone help me understand the pricing of zed? $10 per month- $5 credits for AI credits. This credits can be used for claude code / codex inside zed or should I manage different api keys for codex/claude code?

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?