Sandip Bhattacharya

Many months back when I moved to Bluesky wholeheartedly and deleted everything on Twitter, I made a personal decision for future SM usage - I will make original posts only on Bluesky. I will continue to maintain engagement on Twitter, but restrict myself to replies and simple reposts (without quote)

It is pretty surreal to the endemic corruption inside a country have such an impact on the entire world.

This is hall of fame, “alternative facts” level gold 😭

Some of the weirdest experiences when I came to Canada a few yrs ago, was to find these posters from Kingston to Burlington about various naval battles between Canadians and Americans from the 1800s It always felt unreal given how close Canadians & Americans were. Now, of course, it hits differently

Wealthsimple weekly millionaire promo has got to be one of the most confusing lotto I have ever seen. “reveal” never reveals

I'm not saying it's anyone's responsibility to praise AI usage. I agree with much of the criticism of the tool. But for many, many people, it is a useful tech (I have talked about the ways I find it extraordinarily useful, only to have people here mock me, insisting my lived experience is a lie).

I wouldn't bring this up if it was just one person, but it was multiple people since the beginning of August who mentioned in passing something to the effect of either "I'd spend more time on Bluesky if people weren't so anti-AI" *or* "As much as I hate to be there all the useful AI talk is on X."

Multiple people I know have told me that they love the idea of Bluesky, and want it to succeed, but have abandoned it for X because the use agentic tools in their work and find them incredibly useful, and feel that any mention of their usage here leads to hate and ridicule.

So I get that there's (yet again) a big debate going on about AI on Bluesky. I am turning off replies to this post because I don't want to have notifications flooded one way or the other on this one. I just want to share what I've heard multiple times in just the past two weeks.

okay. I had not realized how good the Indian transliteration input methods are in macos. Just tried the Bangla one and it was perfect!

Decided to switch from Opencode Go to Ollama Cloud even though it is twice as expensive because opencode still insists on serving some models like deepseek 4.1 flash exclusively from outside US

Wow. US model providers keep adding nanny gates, leaving the market wide open to other models.

I use the same AGENTS/CLAUDE.md with Claude code, Codex & Pi, and it annoying how it is always CC which has trouble with following instructions. My feeling is that its system prompt de-prioritizes use instructions way more than other harnesses. Why? I am guessing because they think they know better.

This season of Reacher has been really distracting for me. Seeing him running around my neighborhood parks and my commute route here in Toronto pretending it is Philadelphia, has really made it hard to focus on the story. Last episode he was clearly hanging outside our union station

Sometimes it's easier to show than tell. We're sure this update will help with that. 👀 GitHub CLI now has a repeatable --attach flag that uploads a local image or video. Reference it inline in an issue, pull request, or comment body.

Me, on autopilot, asking ChatGPT "why am i going on having these timeouts. can you check if ChatGPT is down?" 🤦‍♂️

Sigh. I just downgraded to ChatGPT Plus just a week ago. And OpenAI really made it unusable for even mildly serious work with its 5h hours quota. Will probably have to upgrade again.

omg. Codex chrome extension has been broken for days. Problem with the browser use skill. If I use this specific instruction in the extension session it works """ Use browser:control-in-app-browser. Chrome is explicitly selected—not the in-app Browser. Follow the skill’s explicit Chrome flow. """

I have been trying out Fitbit air for a while. And sometimes it is wildly different from Apple Watch which I am also wearing to compare Last night for example: Apple Watch: “You had a terrible night” Google Health: “quality sleep 🎉“ Now who do I believe. All I feel is meh like every day. 🥲

Tried out Opencode Go subscription to see what the fuss was about. Pros: Damn cheap $10/mo Cons: Laughable limits. For GLM 5.3, 5 hour limit of $3, weekly limit of $7.50, and monthly quota of $15 There is no other subscription tier if this is too less.Doesn't seem worth it imo.

Github's postmortem for the August 17 incident has some really wild numbers. I really really doubt any company would seriously try to be a equivalent competitor in this space. The ones blustering will put in a paywall on day 1. github.blog/news-insight...

Very curious about the people who post slop replies to posts even here. Are there some specific tool that people use or is everyone just setting up some adhoc Hermes agent or something?

Tried docker sandbox for codex today. Absolutely pissed to see that it needs a docker login to use it. The FAQ has a complete bs explanation abt why this is needed for "governance" The product is nice. But this rent seeking behavior isnt. This can't be the only solution out there. Still looking.

Interesting times for me when the top of a truncated tweet just says. "This is not going well for Bhattacharya 😬"

I jotted down a few of the patterns I took away from the Blackhat conference talk by OpenAI researches about the hugging face attack. I am still pretty stunned by what happened and how completely unprepared we are for the Mythos level models dropping all over soon. blog.sandipb.net/2026/08/09/s...

Have been scratching my head as to why my Github PR pushes are not triggering actions, asked codex to diagnose my settings. It found out instead that there is an active Github Actions outage, only 15% webhook events are now being supported. www.githubstatus.com/incidents/qc...

TIL that #Codex has allowed its version of the #ClaudeCode AskUserQuestion to be used in the regular mode for a while now, but it has been hidden behind a feature flag. blog.sandipb.net/2026/08/03/c...

Ok, I am not going to look at Boba tea the same way again. 😬 contextualchinese.com/%E6%B3%A2%E9... h/t hachyderm.io/@skinnylatte...

TIL homebrew can manage npm packages for you too. blog.sandipb.net/2026/08/02/h...

While replacing my NAS drive, I wanted to double check using AI. Frontier model GPT 5.6 Sol got confused even after an utility point out the mapping, made a bad assumption and almost made me pull out a good drive from the NAS. AGI is pretty far out.

Was annoyed enough by #Codex using a different mechanism to disable model invocation in skills to write about it. Annoying why I have to do the same thing in three different ways in different agents. blog.sandipb.net/2026/07/15/c...

Really annoyed to just find out that #Codex doesn't support the disable-model-invocation field in skills, and instead wants a codex specific file in the skill to do so. github.com/openai/codex...

Remember how Apple got so much hate for deprecating old hardware when new ones came out? Even though the older ones were good enough for their users? LLM providers are doing the same. Cheaper older opus and sonnet models would be good enough for a lot of ppl. But no, you need to pay more for new

We are the best!* * best in our own house lol