Mitesh B Ashar

29.9K posts

Mitesh B Ashar

Mitesh B Ashar

@iMBA

NOT an MBA.

Kolkata, India Katılım Ağustos 2007
2.6K Takip Edilen2.8K Takipçiler
Mitesh B Ashar retweetledi
Shantanu Goel
Shantanu Goel@shantanugoel·
Introducing rust-bash!! In the age of AI agents, we need a sandboxed shell, unshackled from slow containers. just-bash is a step in the right direction. But rust-bash takes it further. Faster. Use in any language, not just typescript. Framework agnostic, not just vercel AI sdk MCP built in node native Network policy besides execution policies And much more..
English
3
6
32
2.7K
Akshit Singh
Akshit Singh@ssinghakshitt·
@Ra1kshit I haven't really had a good experience with Codex for backend
English
1
0
0
298
Rakshit (chessiro.com)
if you are struggling with claude code rate limits and how expensive it is here is a cheaper and faster way to code. Codex - 20$ plan GLM 5.1- 10$ plan ( for frontend design tasks only) GLM 5 scores well on design arena, and codex is the best for backend work. This is one of the cheapest ways to come close to SOTA performance in under 30$
English
14
1
58
9.5K
Mitesh B Ashar retweetledi
Kyle Mistele 🏴‍☠️
🚨 URGENT 🚨 if you have run npm install / bun install / equivalent in past couple hours run this - if IoC is found your device is cooked, shut it down and get it to infosec/IT folks asap
Kyle Mistele 🏴‍☠️ tweet media
English
35
201
1.9K
413.8K
Mitesh B Ashar
@shantanugoel @MartinShkreli Exactly the same pattern I've been using to build my orchestration system which is also handed with its own IPC tools. All sessions in a project and all it's worktrees can talk to each other.
English
0
0
1
47
Shantanu Goel
Shantanu Goel@shantanugoel·
@MartinShkreli Yeah this one was a bit more complex. But a simple one works as well. The key unlock for me was the pattern to use a main orchestrator and ask it to spawn agents for each experiment/line of thought. Keeps the main agent on track to keep trying
English
1
0
8
1.5K
Martin Shkreli
Martin Shkreli@MartinShkreli·
what is the best tooling for 24-7 inference/agent-driven research? im trying factory but it stops and asks me questions even though i have 'auto' mode on. tbh i think this is an even bigger killer app than LLM chatbots. who else is out there doing it?
English
156
21
591
125.6K
Mitesh B Ashar retweetledi
siddharth ☻
siddharth ☻@siddharthkp·
i'm seeing a new problematic behavior emerge in medium-big companies because of ai somebody notices a performance issue or missing case in a feature they don’t own or maintain instead of creating a ticket, they try to be nice and prompt their way to a PR very nice but…
English
37
31
629
145.2K
Mitesh B Ashar
Mitesh B Ashar@iMBA·
2 weeks back they released their experiment of building computation layers inside transformer. I'd felt this is one of the strong-signal solutions - one that'll help us beyond the scaling limits, to innovate despite of them! Now they've open sourced it! The feeling is stronger!
Christos Tzamos@ChristosTzamos

1/4 Want to build a computer inside a transformer? Given the wide interest in our project, we are releasing the code and the weights so that others can build on our construction.

English
0
0
0
43
Mitesh B Ashar retweetledi
Matt Pocock
Matt Pocock@mattpocockuk·
Every time an LLM says anything to me, I automatically assume it's BS unless it's read a source confirming it NONE of the non-devs I talk to have this instinct
English
194
50
1.2K
59.6K
Mitesh B Ashar
Mitesh B Ashar@iMBA·
@_svs_ Brains washed and meticulously coated with propaganda-protection layers. Programmed with answers for everything: from Aravallis to CARES-accountability to DeMo. you-tell-me-whatever-but-i-have-a-god
English
0
0
0
203
svs 🇮🇳
svs 🇮🇳@_svs_·
I don't think new allegations will have any effect on Modi. Might even increase his aura amongst his supporters. They will certainly not 'end his political career'.
English
4
1
49
3.9K
Akash Mahajan
Akash Mahajan@makash·
Who here has rented a bare metal server? Or an entire rack? How much did it cost per month?
English
5
1
2
611
Mitesh B Ashar
Mitesh B Ashar@iMBA·
@asmah2107 D. Honestly, feelings(rather thought processes) are centered around meta building, coz lately my focus has been in building my own orchestrator and its been building itself with me in HITL mode. If I go back to answering the feelings question, it's exhilarating and liberating!
English
0
0
1
192
Ashutosh Maheshwari
Ashutosh Maheshwari@asmah2107·
Be honest devs in 2026: How much of your daily code is AI-generated? A) 0-25% (purist) B) 25-50% C) 50-75% D) 75%+ (AI is my pair programmer) Reply with your letter + one word on how it feels.
English
19
3
25
7.3K
Mitesh B Ashar retweetledi
Arnav Gupta
Arnav Gupta@championswimmer·
Agentic this, agentic that. And yet..... Claude Code customer support is tagging Boris and Thariq on Twitter. What is Anthropic (and their customer support partners Intercom) doing about this?
English
1
4
83
8.6K
Mitesh B Ashar
Mitesh B Ashar@iMBA·
@championswimmer Totally! This is shabby at best. With Telegram supporting having much more interactivity layers, I was not really expecting "this" shallow integration from Claude code.
English
0
0
0
138
Mitesh B Ashar retweetledi
Nikhil Shahane
Nikhil Shahane@Bingeljell·
You don't know what you don't know... till you read people's docs -- then you know... More stuff I've stolen as I implement Alfred - my AI agent built from scratch. Thank you @shantanugoel - so may nuggets of invaluable information for a first time scrub like me. (Ref the Oxydra docs)
Nikhil Shahane tweet media
English
0
2
5
998
Mitesh B Ashar retweetledi
HELL WALA
HELL WALA@hellwala·
One israeli cooked by an indian news anchor
English
114
4.7K
34.5K
631.6K
pH
pH@pHequals7·
this has been happening way too often with opus 4.6 these days where it confidently spits out an initial version of a fix and then mid way corrects itself saying this won't work too annoying - the model has been nerfed beyond recognition
pH tweet media
English
3
0
11
896