anurag

founding engineer @TraycerAI | built @digitomize

[anurag @anur7ag]

this is insane! opus level intelligence at 90x cheaper output

probably the best price-performance model out there

here’s how it changed my workflow👇

[DeepSeek @deepseek_ai]

🚀 DeepSeek-V4-Flash Official API is now LIVE in public beta!

🔷 We’ve massively upgraded its Agent capabilities—benchmark scores are now far surpassing the V4-Pro-Preview. Check out the massive performance leap below! 👇
🔷 The official V4-Flash now natively supports the Responses API format and is fully adapted for Codex!

Check out the configuration details in our official API docs: https://t.co/smCwQZMeiq

[anurag @anur7ag]

It solves this with durable context, not bigger prompts.

Every transcript stays on disk, so the receiving agent can grep/semantically search past work instead of carrying massive context windows. Artifacts and decisions are shared separately, making agent handoffs essentially lossless.

[anurag @anur7ag]

stacked PR's ❌
stacked Pancakes ✅

[GitHub @github]

Took us a minute, but stacked PRs are now on GitHub 🥞
https://t.co/lT9GGEEDXP

[anurag @anur7ag]

@coniasorous not so much configured only the theme is gruvebox and for git i use lazygit

[anurag @anur7ag]

what if herdr had a desktop app?

[anurag @anur7ag]

@ayush__baliyan

[anurag @anur7ag]

rip my parallel agents

[Theo - t3.gg @theo]

Claude Code has been mostly down for half an hour. Really inconvenient, but I’ll forgive them if we get a reset

[anurag @anur7ag]

classic wrapper behaviour

[shadcn @shadcn]

The new app works with all your AI apps, ChatGPT, Claude, Cursor, T3 but has zero AI features of its own 👀

[anurag @anur7ag]

@walidbelr @thsottiaux goat

[anurag @anur7ag]

honestly, thats a lot of token subsidy

[anurag @anur7ag]

peek before opening agents.

model, branch, worktree, linked PR, last activity- what else would you want to know before jumping in?

[anurag @anur7ag]

@ksjitendra18 On $200, but them reset fells like I’m on infinite money glitch

[anurag @anur7ag]

@walidbelr Damm! How many subs you have or is it the resets?

[anurag @anur7ag]

@AriesTheCoder $14k total spend across claude and codex on a mere $600 sub spend(2xClaude + 1xOpenai)

Thats 23x subsidy!!

[anurag @anur7ag]

@selfdrivenpower glad to know that it is your daily driver. Really want to know more about the Pi issue, want to fix it asap

Will connect in DMs

[anurag @anur7ag]

agents create worktrees like they are free. my disk space disagrees

[anurag @anur7ag]

@madarco you dont have to in a sandbox? 🤔

[anurag @anur7ag]

@Sithira269 yep😬. I heard somwhere that mac needs extra memory space to efficiently manage swap memory that may be one of the reason of sluggish-ishness

[anurag @anur7ag]

@fourlightson its a internal feature we have been testing 😅, not out yet

[anurag @anur7ag]

kept noticing that i was asking agents to do the same kinds of things everyday.

made an automation, so now, every night, it goes through all my chat transcripts, looking for repeated works, and updates existing skills or make new ones.

sounds small, but i've been repeating myself a lot less lately.

[anurag @anur7ag]

your agent picked the wrong approach again? branch it from before it went wrong