Agni Bhattacharyya

Software Engineer | Prev @funnelstory

[jack @jack]

the government of india does not like technologies like bitchat and wants it taken down

[Agni Bhattacharyya @PyAgni]

Claude API reliability sucks, but lowkey feeling good that I'm not always the bottleneck

[Agni Bhattacharyya @PyAgni]

@RohitPaulK Makes sense, can’t disagree. But I’ve been sucked into this fad since past couple months, and only use the IDE for reading/navigating code now. Maybe I should try moving back to Cursor and see how it feels.

[Agni Bhattacharyya @PyAgni]

Code, like steel, is easier to change while it's hot.

- Tiger Style

[Agni Bhattacharyya @PyAgni]

https://t.co/lmhmmL4JbJ

[Agni Bhattacharyya @PyAgni]

Every time my battery dies, Ghostty restores my tabs but all my agent sessions are gone.
Finally got annoyed enough to vibe code a small fix: tab-resume — each tab picks up its most recent claude/grok/opencode/pi session right where the crash left it. One zsh file, no daemon.

[Mitchell Hashimoto @mitchellh]

@poojabnf I read the code

[Mitchell Hashimoto @mitchellh]

The problem with the "if it works who cares what the code looks like" mindset for agentic work is that it assumes the agent has a perfect understanding of "works." Realistically, things are underspecified, agents make bad assumptions, etc.

To be fair, agents are pretty good at unit test coverage. They're pretty bad at designing human experiences (API, CLI flags, etc.), especially cohesive ones for future roadmap plans they may not have visibility into (unless your backlog is perfect and vision fully laid out, which I doubt). They're bad at knowing where performance matters and what type (CPU vs memory tradeoffs). They're bad at where compatibility matters and where it doesn't (and tend to err on the side of preserving it without further guidance). Etc.

Unless you have this ALL specified, you can't possibly claim "it works" without taking a look and thinking about it.

[Agni Bhattacharyya @PyAgni]

@krandiash CARTESIA

[Agni Bhattacharyya @PyAgni]

Great stuff!

[Karan Goel @krandiash]

We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today.

New architectures enable new frontiers for speed and quality.

We're now the only provider to have #1 models for both speaking and listening.

[Agni Bhattacharyya @PyAgni]

The plan mode in @grok build is unreal, idk why no one is talking about it!!

It’s just so interactive and beautifully formatted with rendered colourful markdown, architecture diagrams, imp code snippets, acceptance checklist, INLINE COMMENTS, etc.

cc & oc should take notes!

[Agni Bhattacharyya @PyAgni]

Looks like the only way to change my @Cloudflare account password is via the forgot password flow 😐

[Suhail @Suhail]

My favorite @elonmusk quote that I often send friends:

Do not fear losing. “You will lose,” Musk says. “It will hurt the first fifty times. When you get used to losing, you will play each game with less emotion.” You will be more fearless, take more risks.

[Agni Bhattacharyya @PyAgni]

@asxna @Puneetrai04

[Agni Bhattacharyya @PyAgni]

3rd being the most important, and the most toughest imo

[0xSero @0xSero]

Rules to being a sane person worth listening to:

1. Do not use AI to write your blogs, posts, emails, etc..
2. Do not form relationships with an LLM
3. Avoid shorts like the plague
4. Touch grass
5. Read books
6. Do not insult and humiliate people online
7. Respect your self

[Agni Bhattacharyya @PyAgni]

Hiring without a resume, but with “favourite programming agentic dev session” is such a nice idea 🤯🫡

[Nirant @NirantK]

I'm looking to hire an AI Engineer to pair with me as we work on some of the most ambitious applied problems in 🇮🇳

Couple of things we've done in last 6 months:

1/ Volume: Helped a consumer app scale to 500K sessions/day with identical retention and 30% lower cost with custom memory + implicit caching

2/ Quality: Simulations, Assessment built with the customer which tell what to improve, and not just API errors

We've done production work which others write benchmark-blogs about

3/ Harnesses: We've built and deployed our own sandboxed agents to do analytics on customer logs with bespoke, highly customized skills + memory work

The role is based out of BLR, since we'll visit customer offices, but expect to WFH by default otherwise

Expected pay: 2-4 L/mo depending on how much technical skills x mental acuity,

Independent of seniority, college etc.

[Agni Bhattacharyya @PyAgni]

“same problem as before but 100x harder because people are crossing domains”

Being a back-end dev, I’m experiencing this so much as I’m trying to learn design just so I can write better prompts.

Figma Make, Stitch, Canva AI, or whatever is pretty useless if I don’t have at least a decent understanding of design.

[dax @thdxr]

some thoughts on figuring out how to keep producing good work

[Agni Bhattacharyya @PyAgni]

@0xSero Sure, why not

[Agni Bhattacharyya @PyAgni]

@0xSero @thdxr 🫣🫣