Shardul

817 posts

Shardul

Shardul

@shxrdexe

Software Engineer | Building solutions with code | Shipping web apps, APIs and everything in between | Python, TypeScript, Node, AI Systems

cloud Katılım Aralık 2016
159 Takip Edilen50 Takipçiler
Shardul
Shardul@shxrdexe·
The more I work on production-grade RAG systems, the more I realize RAG is not just a retrieval problem It’s a trust problem And that trust is built through reliable citations and fresh context.
English
0
0
0
7
Tibo
Tibo@thsottiaux·
You can just codex things
English
232
68
1.9K
96.6K
Shardul
Shardul@shxrdexe·
@claudeai Time to apply for another loan to use opus 4.7
English
0
0
0
16
Claude
Claude@claudeai·
Introducing Claude Opus 4.7, our most capable Opus model yet. It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back. You can hand off your hardest work with less supervision.
Claude tweet media
English
4.7K
10K
80.4K
14M
Shardul
Shardul@shxrdexe·
Knowing how to build production grade RAG systems is extremely important these days....
English
0
0
0
14
Meg McNulty #icml2026
Meg McNulty #icml2026@meggmcnulty·
i've been accused of being a deepfake twice this week and it's starting to hurt my feelings
English
5
1
6
586
Shardul
Shardul@shxrdexe·
Life outside claude is sometimes beautiful.
English
0
0
0
31
Shardul
Shardul@shxrdexe·
@robinebers I agree, and to be honest, I don't need the open source models to be as good as the closed ones. I just want them to be good enough for my workflow to run smoothly.
English
0
0
0
105
Robin Ebers • Build Apps With AI
Robin Ebers • Build Apps With AI@robinebers·
closed models will stay ahead and open-source will not win but Gemma 4 matters more than people think and I need you to understand why because less than a year ago, OpenAI had a model called o3-mini-high it was widely accepted to be the best at reasoning and planning but it was VERY expensive no subscriptions, I paid $100-300/day just to code every day but now Gemma 4 is almost at that same level except it's FREE, jailbroken, and runs on your mac/iPhone locally, without a cloud, while you're offline it's not the best model today, but it would have been a year ago genuinely wild to think about this
English
10
1
27
2.3K
Shardul
Shardul@shxrdexe·
@robinebers Cursor is a really good harness but unfortunately quite expensive to use
English
0
0
2
178
Robin Ebers • Build Apps With AI
Robin Ebers • Build Apps With AI@robinebers·
keep looking for a Cursor alternative there still is none still the best tool by a long shot really wish it wasn't, because competition is good
English
29
0
92
8.5K
Shardul
Shardul@shxrdexe·
It took me a while to hit the rate limits. I guess I'll have to build a few more applications to hit the limits again.
Shardul tweet media
Tibo@thsottiaux

Our Codex dashboards are showing increased rate of users hitting rate limits and since we don't fully understand why I have made the cautious decision of resetting the usage limits for all plans. Enjoy. I also wanted to celebrate us finding a pocket of fraudulent accounts that we banned and have helped us regain some compute. The fight against abuse never stops, but it's important to mark the moment and make it a little shared victory.

English
0
0
0
54
Shardul
Shardul@shxrdexe·
I was reading about context engineering and came across a problem a lot of people face. Having multiple MCP servers connected to your AI agent might feel productive, but it can lead to token wastage. Each tool schema easily consumes more than 500 tokens in the background. By the time your model starts thinking, you might have already burned through 50k tokens, just on setup. To fix this, context engines are used. A context engine pulls only what's relevant to the current task. Basically, lazy loading for your AI agent and it makes a bigger difference than you'd think.
English
0
0
0
24
Shardul
Shardul@shxrdexe·
@DeryaTR_ I really thought they would focus on making Opus cheaper with the Opus 4.6 release, but they somehow made it more expensive to use.
English
0
0
2
253
Derya Unutmaz, MD
Derya Unutmaz, MD@DeryaTR_·
Normally, I would greatly praise this & immediately try it, but I am afraid to touch Claude now since I don’t want to consume my weekly max account rates in 5 minutes! There is no point in releasing great features if it becomes prohibitive to use! A bit of a fast & loose move!
Claude@claudeai

Computer use is now in Claude Code. Claude can open your apps, click through your UI, and test what it built, right from the CLI. Now in research preview on Pro and Max plans.

English
64
36
1K
53.7K
Shardul
Shardul@shxrdexe·
Never thought we would be getting open sourced Claude Code before GTA 6.
English
0
0
1
51
Shardul
Shardul@shxrdexe·
@Amank1412 Anthropic might become profitable soon. I don't see that happening for OpenAI anytime soon.
English
0
0
1
311
Aman
Aman@Amank1412·
I don’t get what investors see in OpenAI/Anthropic. Not profitable. Burning cash. Yet valuations keep going up. The math isn’t mathing.
English
112
7
222
23.1K
toucan
toucan@distributionat·
It's really surprising that ChatGPT's search is superior to Gemini's. Like what are you doing Google?
English
20
5
242
22.1K
Shardul
Shardul@shxrdexe·
Google's TurboQuant release might have an indirect connection with their Apple/Siri collaboration deal. Apple wants to power Siri for billions of users, the next logical step isn't just running models on device, it's letting users control their entire phone through them. I think Google is quietly laying the groundwork for that. Or maybe I'm just overthinking.
English
0
0
0
40
Shardul
Shardul@shxrdexe·
@mSanterre It's also more expensive to use than claude code.
English
0
0
1
150
Shardul
Shardul@shxrdexe·
@meggmcnulty @Pirat_Nation I agree, the RTX 5090 was a letdown for me. Not much better than the 4090, and way too expensive.
English
0
0
1
23
Meg McNulty #icml2026
Meg McNulty #icml2026@meggmcnulty·
@Pirat_Nation Big jumps on paper, but the real story will be how it performs in actual games. Drivers, optimization, and memory tuning usually decide more than raw specs.
English
1
0
1
1K
Pirat_Nation 🔴
Pirat_Nation 🔴@Pirat_Nation·
NVIDIA's next RTX 60 series graphics new rumors >30-35% better rasterization performance and >Up double the ray tracing speed compared to the RTX 50 series. >The RTX 6090, may have over 24,000 CUDA cores, 32GB of fast GDDR7 memory, and use a new 3nm process. Nvidia hasn´t confirmed this
Pirat_Nation 🔴 tweet mediaPirat_Nation 🔴 tweet media
English
191
93
1.8K
156.7K
Shardul
Shardul@shxrdexe·
@OpenRouter @Alibaba_Qwen Didn't they just release omni? Alibaba is trying to keep up with Anthropic announcements it seems.
English
3
0
23
5K
OpenRouter
OpenRouter@OpenRouter·
Qwen 3.6 Plus Preview from @Alibaba_Qwen is live now for free for a limited time on OpenRouter! During this free period, prompts and completions will be collected and may be used to improve the model.
OpenRouter tweet media
English
35
129
1.5K
269.3K
Shardul
Shardul@shxrdexe·
@claudeai Anthropic is gearing up to make me bankrupt. Nice.
English
0
0
0
14
Claude
Claude@claudeai·
Computer use is now in Claude Code. Claude can open your apps, click through your UI, and test what it built, right from the CLI. Now in research preview on Pro and Max plans.
English
2.5K
4.7K
58.9K
16.2M

Keşfet

@thsottiaux @claudeai @meggmcnulty @robinebers @DeryaTR_ @Amank1412 @elonmusk @BarackObama