27 | Curious personš½| Private Equity Tech Consultant | Bentley U ā21 | š²š±
Boston, MA
Joined April 2012
- Tweets3.3K
- Following2K
- Followers1.4K
- Likes97K
Pinned Tweet
Once a day I regret going the Investment Banking and PE route and not doing CS in college, but still taught myself code and still won a Google hackathon. Proud of myselfš¤ shout out @GoogleDeepMind @GoogleAIStudio shoutout @NanoBanana for being elite.
decided last minute to build for Googleās Nano Banana Hackathon. Had fun with this one. Not bad for 15hours of codingšāāļø. gemini is elite. found a simple architecture that allows for consistency & coherence over long stories. Just gotta add veo3 now kumayiri.com
with no image model you would expect claude to struggle with vision when looking at things, yet it genuinely may have the best vision. i used to have gemini as best vision but opus may have taken lead
revolutionary for the normies, theyāre about to understand more
A few big upgrades for Chrome users today! Weāre bringing the power of computer use to anyone who uses Chrome.
First up: ChatGPT can now use your tabs in Chrome to make answers more useful and get things done for you.
If youāre in Chat, it can look across your open tabs and gather info.
If youāre in Work, ChatGPT can control your Chrome tabs to get things, like sending messages, filling out forms, or downloading files. It can even help you organize and group that pile of tabs you have open.
a lot of people expected this to start happening much sooner but only a matter of time if we keep being stupid. weāve been lucky that high gas prices is main discussed front issue in the USā¦for now
We currently see no signs from Russia that it wants to end the war. We see no basic steps toward de-escalation. Everyone is talking about diesel and forgetting about grain. In a month or two, everyone will start talking about an agricultural crisis. There wonāt be enough food for countries across Africa, some Asian countries, and the Middle East, because we export millions of tons of grain to these very markets. And now Russia is blocking the sea.
The United States isnāt feeling the effects of this, which is why the focus is only on energy. This is another challenge. Itās hunger. And it can affect many countries. And we need to think about everything and everyone, because we all live on this planet. We can prevent many different crises before they even begin and avert the challenges they bring.
From remarks at the Warsaw Security Forum. (2)
have to hand it to the claude code team fr with projects, not saying codex canāt already do something like this because it can, but on codex you have to know what youāre doing more; projects is very targeted marketing and its own solution. they know the dev communityās pain points, these two lines tell you exactly the product quality they are going for with projects
man the tables have turned lol wild it took this long to get this in codex, claudeās had this forever, even @julesagent BY GOOGLE had this last year
been living a life of too many self inflicted wounds for the past few years, gotta up the discipline
moved up and finally broke top 100 at 55th. itās crazy because just a few weeks ago, literally week of September 6th, this score would have placed me top 5
Probably evidence of how much these new models are speeding up research. this ARC AGI competition started march and no one broke 7 points until this month, and now at the end of the month, with a September 30th deadline, since astra since opus 5.5, the new high score is 45
ggs @tufalabs š«”
this has been my best personal benchmark for all models. Testing their intelligence and ability to be creative in problem solving to come up with ānovelā solutions. i had given astra high another 2weeks to improve its ARC AGI hackathon score, it frustratingly topped off 2.84 before i had it stop trying.
Last night i started a new repo for opus 5.5 high, woke up to it scoring 4.31, ranking me 142nd of 3,436š¤, deadline is in 2 days so weāll see if opus can improve in that timeš¤
to be fair to astra, maybe i was holding its hands too much, and was an unnecessary stupid human in the loop messing it up, i sent astra over 50 prompts through the two weeks, iāve only sent opus one prompt so far
this has been my best personal benchmark for all models. Testing their intelligence and ability to be creative in problem solving to come up with ānovelā solutions. i had given astra high another 2weeks to improve its ARC AGI hackathon score, it frustratingly topped off 2.84 before i had it stop trying.
Last night i started a new repo for opus 5.5 high, woke up to it scoring 4.31, ranking me 142nd of 3,436š¤, deadline is in 2 days so weāll see if opus can improve in that timeš¤
to be fair to astra, maybe i was holding its hands too much, and was an unnecessary stupid human in the loop messing it up, i sent astra over 50 prompts through the two weeks, iāve only sent opus one prompt so far
Past 2 months i tried to see if Sol could place me on the leaderboard for ARC AGI 3 comp on @kaggle , got 0.0 for 2months straight lol. Past 2 days, had Astra max come in, one prompt, immediate ranking. nothing crazy, just 1.59 w Qwen3.8, but evidence of a jump in intelligenceš¤
same, fully agree. as great as codex and astra are, OpenAI DESPERATELY needs an answer to Opusās creativity and visual prowess. Opusās creativity shows up even in its reasoning of complex things, math and science, in ways that astra just canāt compete with right now, it has shocked me a lot as one of anthropicās top haters
Shout out to opus 5.5 for the redesign on my soulnests.com app, well done, it created a nice demo video of its work too
