@itsOmSarraf_ @Jitesh_117 You work on sarvam samvaad. Lol i also have a product called samvaad.
English
Atharva Verma
433 posts
Atharva Verma
@atharva_again
building cool stuff. checkout https://t.co/IUq7AIk2oO and https://t.co/U8y5BQ4VGz
Katılım Nisan 2022
94 Takip Edilen11 Takipçiler
na bachhon mein internship crack karne ki haisiyat hai na hi college mein NOC dene ki neeyat
VG🌪️@HelloVyom
Why do we never see Indian students with sooo much work experience while students in the west manage to do 5-6 high quality internships in tech even before graduating? Look at this -> Meta, Stripe, Rippling, Snowflake, Questrade, Samsung ☠️
Eesti
lmao
> dotnet run
Luke Parker@LukeParkerDev
clippy is back... kinda. opencode as a desktop pet
English
@AmanHasNoName_2 improper planning. atleast dmrc built that bridge between dhula kuan and durgabai deshmukh.
English
??? Isnt this basic??
Why would anyone step outside the metro station to change lines ???
Roads of Mumbai@RoadsOfMumbai
In Delhi, you can interchange from Yellow Line to Violet Line to Pink Line without stepping a foot outside the metro station premises all on a single ticket! Next level execution 🤯
English
@sushantstwt @devfolio that's reall cool. can you make inviting team members easier? i usually have the same team for hackathons, and manually copying the hackathon url and team code is tedious. you can look at how unstop does it, it's better.
English
@ryanvogel how is this new? i have been using omo from the beginning and it handles this from the start.
English
"yeah you can copy my work but change it up a bit so it doesn't look like you copied"
ClaudeDevs@ClaudeDevs
How do you keep Claude working until the job is done? Claude Code helps with this in a few ways, including one we shipped recently: /goal.
English
@samlambert just yesterday i set minimum age release and no exotic builds in my npmrc and global pnpm and bun config as well as uv config. cant take any risks.
English
We’re hiring Founding Engineer 🚀
Work directly with me and our incredible team to build the core product experience from the ground up - Both long-running agent itself and everything needed to scale fast.
We’re looking for someone who:
- thrives in ambiguity
- has strong product intuition and cares deeply about craft
- ships fast, takes ownership, and relentlessly raises the quality bar
- loves talking to customers and building from real user pain points
If you’ve ever wanted to help define a category from day one - or know someone exceptional who would - DM me. Would love to chat.
English
@liadyosef @idosal1 @Cloudflare Can i ask, why an mcp when the agent can just clone the entire repo and understand it?
English
GitMCP went dark today, and then came back to life.
Being a free, open source project, and more popular than ever - the bills reached over $14,000 this month.
It was automatically shut off, and @idosal1 and I were about to take it down.
The @Cloudflare team moved fast to bring us under their Project Alexandria program - now fully supporting GitMCP as an open source project. It stays online and free for everyone using it.
Massive thanks to Valentina, @irvinebroque and the @Cloudflare team. This is what real support for open source looks like. 👏
English
Whole day with Pi agent just used 10% of my $20 Codex weekly quota
All gpt 5.5, various thinking levels
I'm now 3 hours away from weekly reset and still have 42% left!
For my typical 4-6h/day session:
Codex Desktop - 20-30% a day, 1-2 days/week
Pi - 10% a day, 4-5 days/week?
My switch to Pi is really looking good so far
raymel 👋@pseudokid
Pi Agent vs OpenCode token usage A lot of people recommended Pi Agent so I decided to check Pi Agent took 1.1k tokens in first turn OpenCode took 11.5k Setup: 1) Trimmed OpenCode (from usual 30k first turn to 11.5k) - 0 MCPs - 2 lightweight plugins (opencode-env-protect and openslimedit) - 8k char AGENTS.md - 11585 deepseek-v4-flash input tokens - for $0.0016 2) Vanilla Pi - 0 MCPs - 0 system prompt - 8k char AGENTS.md - 1114 kimi-k2.6 input tokens - for $0.0008 I think using better models with capped tokens per turn can keep usage nearly the same as uncontrolled DeepSeek V4 Flash? The challenge now is finding the sweet spot. Can't cap tokens if quality drops. We'll see. Video below - OpenCode vs Pi side by side "say hi back" first turns test, with OpenCode Go usage for each
English
Pi Agent vs OpenCode token usage
A lot of people recommended Pi Agent so I decided to check
Pi Agent took 1.1k tokens in first turn
OpenCode took 11.5k
Setup:
1) Trimmed OpenCode (from usual 30k first turn to 11.5k)
- 0 MCPs
- 2 lightweight plugins (opencode-env-protect and openslimedit)
- 8k char AGENTS.md
- 11585 deepseek-v4-flash input tokens
- for $0.0016
2) Vanilla Pi
- 0 MCPs
- 0 system prompt
- 8k char AGENTS.md
- 1114 kimi-k2.6 input tokens
- for $0.0008
I think using better models with capped tokens per turn can keep usage nearly the same as uncontrolled DeepSeek V4 Flash?
The challenge now is finding the sweet spot. Can't cap tokens if quality drops.
We'll see.
Video below - OpenCode vs Pi side by side "say hi back" first turns test, with OpenCode Go usage for each
raymel 👋@pseudokid
Sorry guys, I'm now a Codex user I love my OpenCode, but it eats my ChatGPT limits thrice than Codex Desktop will ever be in a day I still love TUIs but GUIs aren't bad either
English
@trq212 But can you edit an html file without using claude code? I mean ik you obviously can since you're a dev, but can other non devs do that? Whereas something like markdown is pretty trivial to use.
English
HTML is the new markdown.
I've stopped writing markdown files for almost everything and switched to using Claude Code to generate HTML for me. This is why.
Thariq@trq212
English
Engram is finally done
it's a Rust memory backend for LLM agents that sits between your agent and its conversation history
made up of three memory layers:
- short-term (Redis)
- long-term semantic search (LanceDB)
- core pinned facts
full REST API, token budgeting per request, pair preserving trim, background workers, Prometheus, and Swagger
raw numbers from my laptop:
- 100 msgs context assembly: 280 µs in-memory, 22 ms real stores
- e2e throughput: 64k msg/s
- token savings vs full dump: 40%
repo link in comments
English
logging is the most underrated feature in any system. you never think about it until you need it, and when you need it, nothing else matters.
English
@pratyush_r8 @district_india cdg's logo. heart with eyes. so simple. but so beautiful. (5/5)
English
@pratyush_r8 @district_india the dune sofa - made to facilitate rest and conversations (4/5)
English
Help us improve our design taste!
We're hiring a Visual Design Intern. 50k+ stipend, Bangalore-based.
Comment 5 apps, softwares, hardware, food, products - anything you feel is tastefully designed.
And a one-liner on why (mandatory).
That's it. That's your Round 1.
If your taste checks out, we'll DM you with the next steps.
The Filter:
1. Role: Visual design, focused on creating bespoke UI, animations & interactions.
2. Application: Comment below. That's your only task.
3. The Signal: We're not looking for @lovable heroes. We want people who notice the weight of a scroll, the pause before a transition, the reason a menu feels right.
If you feel claude won't eat your job but make you 420x - you're our person.
Pay: Min 50k. No cap for tasty talent.
Location: Preferably Bangalore. Remote for those who really really really can't make it and we both love each other.
Deadline: none, always open to hiring great talent.
P.S. To know out taste, go on @ThineAI . The image below is a terrible AI generated shit. Please don't consider that as a benchmark :P
Thnx and see you in comments
English