Logo
Picture

Arni Mittal @arnimittal

behavioural psych, healthcare, ai verification tryzerocheck.com Internet Joined July 2012
  • Tweets
    187
  • Followers
    437
  • Following
    316
  • Likes
    7K
Arni Mittal retweeted
GrantSlatton Profile Picture

Grant Slatton @GrantSlatton

a month ago

@xeophon Another hack I use for open-ended goals is: /goal 1. Make a report file 2. Generate 100 ideas to improve <benchmark> 3. Try them all 4. Only accept those that <constraints> 5. After trying all, if any accepted, go back to step 2, else exit

Arni Mittal retweeted
astro_reid Profile Picture

Reid Wiseman @astro_reid

3 months ago

Only one chance in this lifetime… Like watching sunset at the beach from the most foreign seat in the cosmos, I couldn’t resist a cell phone video of Earthset. You can hear the shutter on the Nikon as @Astro_Christina is hammering away on 3-shot brackets and capturing those exceptional Earthset photos through the 400mm lens. @AstroVicGlover was in window 3 watching with @Astro_Jeremy next to him. I could barely see the Moon through the docking hatch window but the iPhone was the perfect size to catch the view…this is uncropped, uncut with 8x zoom which is quite comparable to the view of the human eye. Enjoy.

Arni Mittal retweeted
chetaslua Profile Picture

Chetaslua @chetaslua

3 months ago

@HarshithLucky3 @OpenAI @sama holy shit

Arni Mittal retweeted
karpathy Profile Picture

Andrej Karpathy @karpathy

4 months ago

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes, today I measured that the leaderboard's "Time to GPT-2" drops from 2.02 hours to 1.80 hours (~11% improvement), this will be the new leaderboard entry. So yes, these are real improvements and they make an actual difference. I am mildly surprised that my very first naive attempt already worked this well on top of what I thought was already a fairly manually well-tuned project. This is a first for me because I am very used to doing the iterative optimization of neural network training manually. You come up with ideas, you implement them, you check if they work (better validation loss), you come up with new ideas based on that, you read some papers for inspiration, etc etc. This is the bread and butter of what I do daily for 2 decades. Seeing the agent do this entire workflow end-to-end and all by itself as it worked through approx. 700 changes autonomously is wild. It really looked at the sequence of results of experiments and used that to plan the next ones. It's not novel, ground-breaking "research" (yet), but all the adjustments are "real", I didn't find them manually previously, and they stack up and actually improved nanochat. Among the bigger things e.g.: - It noticed an oversight that my parameterless QKnorm didn't have a scaler multiplier attached, so my attention was too diffuse. The agent found multipliers to sharpen it, pointing to future work. - It found that the Value Embeddings really like regularization and I wasn't applying any (oops). - It found that my banded attention was too conservative (i forgot to tune it). - It found that AdamW betas were all messed up. - It tuned the weight decay schedule. - It tuned the network initialization. This is on top of all the tuning I've already done over a good amount of time. The exact commit is here, from this "round 1" of autoresearch. I am going to kick off "round 2", and in parallel I am looking at how multiple agents can collaborate to unlock parallelism. github.com/karpathy/nanoc… All LLM frontier labs will do this. It's the final boss battle. It's a lot more complex at scale of course - you don't just have a single train. py file to tune. But doing it is "just engineering" and it's going to work. You spin up a swarm of agents, you have them collaborate to tune smaller models, you promote the most promising ideas to increasingly larger scales, and humans (optionally) contribute on the edges. And more generally, *any* metric you care about that is reasonably efficient to evaluate (or that has more efficient proxy metrics such as training a smaller network) can be autoresearched by an agent swarm. It's worth thinking about whether your problem falls into this bucket too.

Arni Mittal retweeted
vasuman Profile Picture

vas @vasuman

4 months ago

Somewhere out there is a guy who uses Notion, Superhuman, OpenClaw on a Mac Mini, Raycast, a mechanical keyboard ($400), Wispr Flow, and gets nothing done every day

Arni Mittal retweeted
garrytan Profile Picture

Garry Tan @garrytan

5 months ago

Is there any way I can connect Claude Code to use my browser and iOS simulator so it can find mobile css bugs and make things responsive on tablet, web, and mobile automatically?

arnimittal Profile Picture

Arni Mittal @arnimittal

6 months ago

youtube.com/watch?v=FIOqk4…

Arni Mittal retweeted
Quartr_App Profile Picture

Quartr @Quartr_App

8 months ago

$UBER Q3 2025 "Uber's growth kicked into high gear in Q3, marking one of the largest trip-volume increases in the company's history" – Dara Khosrowshahi, CEO MAPCs* +17% Trips +22% Gross Bookings +21% Revenue +20% *Mobility +20% *Delivery +29% *Freight —% EBIT +5% *marg. 8.3% (9.5) FCF +6% *Monthly Active Platform Consumers

Arni Mittal retweeted
Quartr_App Profile Picture

Quartr @Quartr_App

8 months ago

Arni Mittal retweeted
Timb03 Profile Picture

Tim Bennetto @Timb03

9 months ago

Pallyy has NEVER been below 5% churn. And it's not by listening to cancel reasons. Combo of better onboarding + activation emails and improvements to the product, particularly around the first actions a user would take.

Arni Mittal retweeted
mintlify Profile Picture

Mintlify @mintlify

10 months ago

We just killed traditional 404 pages When someone hits a dead link → AI agent reads the path → suggests semantically similar pages Shoutout to @masonwilliams for the suggestion!

Arni Mittal retweeted
amasad Profile Picture

Amjad Masad @amasad

10 months ago

Automated Testing. While building your app, Agent 3 will periodically test it using a browser, generating a report and fixing the issues it finds. Our proprietary testing system is 3x faster and 10x more cost effective than state-of-the-art Computer Use Models.

Arni Mittal retweeted
TurnerNovak Profile Picture

Turner Novak 🍌🧢 @TurnerNovak

10 months ago

prepping for the great lock-in

Arni Mittal retweeted
paulg Profile Picture

Paul Graham @paulg

10 months ago

@sofipriors Hire someone iff the lack of that person is the main thing holding back your revenue growth.

Arni Mittal retweeted
bayeslord Profile Picture

bayes @bayeslord

10 months ago

if steve jobs were alive employee: “hiii steve here’s our final apple vision pro prototype! it’s made from recycled metal!” jobs: *uses it for 5 seconds* jobs: “WHAT THE FUCK IS THIS WHY DOES IT WEIGH 200 POUNDS AND WHY CANT I USE MULTIPLE MONITORS? YOU’RE ALL FUCKING FIRED”

arnimittal Profile Picture

Arni Mittal @arnimittal

11 months ago

@marty_kausas 🔥🔥

Arni Mittal retweeted
spottedinprod Profile Picture

SIP @spottedinprod

11 months ago

alltrails · conditions by @AllTrails

Arni Mittal retweeted
adcock_brett Profile Picture

Brett Adcock @adcock_brett

11 months ago

Anything else Noam?

polynoamial Profile Picture

Noam Brown @polynoamial

11 months ago

@kimmonismus Does it still work if you raise the table 6 inches?

Arni Mittal retweeted
0xgaut Profile Picture

gaut @0xgaut

11 months ago

"build me an app that will generate $100M ARR. don't make mistakes. don't use emdashes"

Arni Mittal retweeted
creatine_cycle Profile Picture

atlas @creatine_cycle

11 months ago

here is what happens when you take creatine: - 5gs: bigger muscles - 15gs: bigger brain - 70gs: replace sleep - 88gs: remote viewing - 120gs: agentic workflows

Trends for United States

Happy 4th of July
#4thJuly
Beyoncé
Donk
Celebrating 250
MORNING DEW
Good Saturday
For 250
Here's to 250
God Bless the USA
Happy Birthday USA
God is the Lord
#USA250
#TicketToHeavenFinalEP
Freedom
Terry
Rybakina
#Caturday
#GodBlessUSA
Mount Rushmore

You might like

growth @ euc, prev startmate & atlassian
Niel Reitmann
@NielReitmann
974 Followers
@487studio / @superflowstudio (making websites) + @blackbirdvc (backing young people)
Anubhav Ghosh
@anubhavg8
1K Followers
in a gradual awakening
abby
@abbyshenn
737 Followers
growth @chronicle_hq 𓇢𓆸
Renee Zhang
@rene3zhang
754 Followers
Championing for climate and DEI through tech, community and theatre | she/her
Eezu Tan
@EezuTan
519 Followers
Founder @migratemate
Mihailo Bozic
@MihailoBozic
1K Followers
The most basic thing about me is my bio: Tech, People, Sports (MUFC, LAA, GSW, Nadal, F1), Dessert, Anime, Manga, Open world games, History and Economics
Yash
@nextdooryash
334 Followers