Varun Godbole
@VarunGodbole
Followers
903
Following
10K
Media
4
Statuses
819
ex-@GoogleDeepMind. Used to work on Gemini. Taking a sabbatical to explore my curiosity.
New York, USA
Joined March 2013
When you talk to AI, you’re effectively conversing with rocks that we’ve tricked into thinking
845
870
12K
He moves through a realm not made for him, where order has replaced nature. A world that mirrors the shape of creation but improperly. The mechanisms hum like wind streams without souls, precise but hollow, lacking spirit. Where am I, he wonders. Sometimes, I understand him.
146
1K
12K
remember when we had to copy paste code from chatgpt on a browser to the ide manually
283
92
3K
most people who say they work 80-100 hours a week i interpret it as all fluff (investment banking, PMs) most of the time is spent sitting around waiting for things to happen, sending slacks and emails there are rare people who actually work 80-100 hours a week. they look like
475
296
10K
My friend just wrote a book!
My book Conscious Accomplishment - How to Use Personal Achievement For Spiritual Growth is now available! For decades I tried to achieve my way to happiness. When this stopped working, I was forced to turn my focus inward. Pretty quickly, I began to realize that working on my
0
0
2
I resonated with pretty much every single thing Tom's posted here about life transitions. I wish I'd seen this a year ago. But even if I had, I'm not sure I'd have actually understood it 😅
I wrote a listicle aggregating some of the most common 10 findings from the last 10 years of speaking to people about life transitions and personal evolution.
1
0
1
Does anyone have suggestions on the best way to feed PDFs, Word, PowerPoint documents into Claude Code or Codex? Is there an OSS implementation that intelligently decides whether one of these documents should be represented as text or multimodal tokens, or both?
1
0
0
What's the best LLM on the market for manipulating CSVs? For example, spreadsheets of various financial information, inventory, etc? And by LLM, I mean without the use of tools.
0
0
1
Does anyone have pointers to a resource that explains to non tech normies the difference between a system instruction and a user turn?
1
0
0
when an AI researcher tells you they're "working all day" they're not really working. they just take their laptop to the bathroom to check how their runs are doing at 2 am but its not actually work. just a severe gambling addiction
28
35
866
@trq212 @dexhorthy @hellovai @danshipper @steipete @PreslavMihaylov haha at this point I'm just tagging people I follow that seem to use Claude Code a lot. I'm seriously tempted to make some sort of lightweight "compiler" for this. But it also feels "wrong" somehow.
4
0
0
2
0
0
@trq212 @dexhorthy @hellovai @danshipper have you run into this? If so, how do you deal with this? Or do you just have an instruction in your https://t.co/5GHMlzfHWV to tell Claude to always read a specific list of files in the repo? That feels like a waste of tokens though....
2
0
0
My https://t.co/5GHMlzfHWV is becoming unwieldy with lots of sections. I often need to reuse parts of it between repos. IIUC https://t.co/5GHMlzfHWV doesn't support import statements. I'd be a game-changer for organization. Ditto for a template system to go with it.
2
0
2
Context windows are like RAM. You can never have enough! I want a 10M context window LLM!!
1
0
2
If the security for your LLM-powered system relies entirely on this assumption: "The attacker can never make the LLM output [X]" (for some particular 'dangerous' value of X) ...then I regret to inform you that you are in for a very bad time.
2
3
13
People leaving regular companies: Time for a change! Excited for my next chapter! People leaving AI companies: I have gazed into the endless night and there are shapes out there. We must be kind to one another. I am moving on to study philosophy.
96
348
6K