Deepanshu Thakur

@ProtonVPN Well I'm a man of my words @ProtonVPN

@DeepanshuMSA The simple answer is no.

YOUR ISP LOGS YOUR TRAFFIC, USE A VPN YOUR ISP LOGS YOUR TRAFFIC, USE A VPN YOUR ISP LOGS YOUR TRAFFIC, USE A VPN YOUR ISP LOGS YOUR TRAFFIC, USE A VPN YOUR ISP LOGS YOUR TRAFFIC, USE A VPN YOUR ISP LOGS YOUR TRAFFIC, USE A VPN YOUR ISP LOGS YOUR TRAFFIC, USE A VPN

@ProtonVPN That's fine. I trust ProtonVPN in that regards however even with the VPN enabled, technically my ISP can still see the websites I'm visiting, right? Apologies if I'm asking a basic question.

If we're talking about the ABILITY to see, then any VPN can technically see your traffic, at least in real time. A VPN essentially replaces your ISP in that regard, but the major difference is that some VPNs (like ours) don't actually keep any records of that traffic, or monetize it in any way.

@thsottiaux Hey @thsottiaux I recently transitioned from being a Codex Pro developer user to a Codex Pro Business app user. I've been trying to use the Codex app for my computer-use tasks, but it keeps running into issues related to being unable to browse websites within the app.

Unfortunately looking at the demos today, the next few weeks of Codex will be eventful and require many clicks of the update button

@pati_marins64 Hello @pati_marins64 I generally don't write articles or posts but what's your take on this piece? deepanshu2017.medium.com/us-israel-iran…

War Causes $3.2 Trillion Losses in Global Markets and pressure US-Israel The United States is deploying its B-52 bombers to attack Iran, along with more B-2s, in a move that appears aimed at devastating Iran and forcing surrender. Today, missiles and drones continued hitting Israel and U.S. bases, but with one detail: almost no reaction from air defenses, which are likely being conserved to protect assets like airports, energy infrastructure, and military industry. With daily costs potentially reaching $10 billion and, according to The Economic Times, losses of $3.2 trillion in global markets already reported in the first 48 hours of war, this is generating more pressure to end the conflict. Israel's central bank governor has announced estimated losses of $17–20 billion. If Israel's losses are that high, how much are the American ones? The order is to intensify efforts to end the war quickly. However, the waves of bombings hitting Iran do not seem to have that effect. Iran's entire military doctrine was built around such scenarios, which is why it has such an extensive network of multi-kilometer underground bases. I see no possibility of the war ending in the coming days for several reasons, including: 1. In addition to their naval strength of hundreds of remaining assets, Iran still maintains a sustained rate of fire, deploying dozens of missiles and hundreds of drones every daqui ; 2. The decentralization of operations. For example, anti-ship missiles are operated by four forces: IRGC, navy, army, and aerospace forces. The same decentralization applies to drones and other military assets, creating resilience; 3. The belief that the US and Israel cannot sustain a long war. Iranians know that, although missile attacks have been reduced, current waves are enough to continue neutralizing most U.S. bases in the region, maintain economic chaos in Israel and Arab countries, and drive losses in global markets; 4. Iranians are certain that no matter how many bombs fall, they will emerge victorious, since there is no way to destroy their nuclear program except by themselves, and the war has virtually suppressed internal opposition movements. 5. The attacks have not sparked internal guerrilla movements or any uprising against the government. On the contrary, the level of violence has increased support for the regime. It's that old story: if the dose of the medicine is too high, it can have the opposite effect. Ironically, Iran is heading toward a scenario similar to Ukraine's, where everyone knows there's no way to win against the Russians, but the cost keeps getting higher. In fact, when I read Pentagon reports, they remind me a lot of the Russian MoD's in the early days of the Ukraine invasion. Iranian missiles and drones launched today face almost no resistance from air defenses already depleted in many areas, the Strait of Hormuz remains closed, Iranian naval forces are nearly intact, markets are in panic, and Iran still has dozens of underground military fortresses loaded with missiles and equipment. Anyone who thinks the US and Israel can sustain this war for weeks in this scenario is very optimistic.

@EndGameWW3 That's nice and bold! And tbh, that makes this page even more trustworthy and interesting to follow. A page that isn't biased towards earning or using words to optimise reach. I just turned on notifications for your posts.

Let me tell the new followers about myself, I don't make any money off of X not one cent, I don't worry about analytics, or how many followers I have. I say what I want to say with 0 fucks given, you want to follow great you don't great. If you're a dickhead I will just simply tell you to eat shit and block you. My own personal freedom with this account...

@testingcatalog Oh damn! This is so cool. How did you get these merchs in the first place? What did I miss? (I was using Codex 5.3 and have been in my shell coding ever since the launch)

It is worth testing Codex ❤️

@VraserX @arcprize Aged like a fine wine. Google announced 84%+ in Arc AGI 2 today

@arcprize This benchmark will be saturated in 2026.

Today, we're announcing a preview of ARC-AGI-3, the Interactive Reasoning Benchmark with the widest gap between easy for humans and hard for AI We’re releasing: * 3 games (environments) * $10K agent contest * AI agents API Starting scores - Frontier AI: 0%, Humans: 100%

@kieranklaassen I started using it recently, it just eats main context when I'm reviewing a PR and most of the times consumes 100% of the main content even before the review is completed. It's an amazing plugin though! I see a lot of work has gone into building this! Appreciated the work!

new release [2.31.0] - 2026-02-08 lots of good stuff and less tokens in context! github.com/EveryInc/compo…

@matt_elevenlabs 11B

We just raised $500M at an 11B valuation 🎉 To celebrate, we’re giving away 1,000 free credits so you can test our platform. For the next 6 hours, comment “11B” below and we’ll DM you the credits (must follow) 👇

@sama @elonmusk is that how is it going to end?

Really excited to get Elon under oath in a few months, Christmas in April!

@Yuchenj_UW @Austen @gdb Who does that? I mean I have never seen anyone benefiting from those diaries ever (neither in movies nor in real life )😅

@Austen @gdb luckily, I don’t do diaries.

I have great respect for Elon, but the way he cherry-picked from my personal journal is beyond dishonest. Elon and we had agreed a for-profit was the next step for OpenAI's mission. The context shows these snippets were actually about whether to accept Elon's draconian terms.

@rauchg @reactjs @nextjs @shuding I read this, tried skills the very first time with cc, asked Claude to apply these best practices and since then my application got super charged, never thought it could perform so well! This is just Wow!!

We're encapsulating all our knowledge of @reactjs & @nextjs frontend optimization into a set of reusable skills for agents. This is a 10+ years of experience from the likes of @shuding, distilled for the benefit of every Ralph

Ever since I started using Claude Code Native on Windows PowerShell, I started seeing these temp files in my working directory (sometimes up to 100s of them) @trq212 @bcherny

@burkov @trq212 I know you have said countless times you don't do that however I agree with this post for the last 2-3 days Claude Opus has performed so bad that I am using Codex now with OpenCode and I'm on $200 Max plan for Claude

I was just about to post that as well! I'm sure Anthropic cheats and serves a weaker model when overloaded or needs GPUs for something urgent. The last two days, I feel like I'm talking to a retard from 2024.

@nikitabier @steipete @instagram Savage 😂

@instagram I’m glad you shared this on X, because no one would see it on Threads.

We fixed an issue that let an external party request password reset emails for some people. There was no breach of our systems and your Instagram accounts are secure. You can ignore those emails — sorry for any confusion.

@thdxr Wow!! Because of all the controversy I tried OpenCode (I'm Claude Code Max 20x and GPT Pro subscriber), after using OC for few hours I have switched away from Codex to OC for sure. CC is still needed for the Opus model but man I didn't know I was missing so much!

we are working with openai to allow codex users to benefit from their subscription directly within OpenCode

@bcherny Love it! Thank you. You are the best @bcherny

@DeepanshuMSA CLAUDE_CODE_FILE_READ_MAX_OUTPUT_TOKENS=12345 claude

Great tip for bigger codebases

@nateberkopec Finally! Someone said it! I need observability and I often steer Claude in real time. One workflow that has actually worked for me after Ralph is, having prd.json file with different feature names and then using Subagents to implement a group of small tasks at once

After a few hours of ralphing, here's my thoughts: 1. Having zero observability on the running agent is rough. No way to steer, check in, notice stalling. 2. You don't find out that your spec was shit until much later. Not sure this fits the way I work