Andrew Gao
@itsandrewgao
Followers
31,076
Following
1,963
Media
1,739
Statuses
7,962
follow for gpt4-o updates! @stanford ; prev @LangChainAI @nomic_ai ; Z Fellow 🇺🇸
San Francisco, CA
Joined October 2020
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
#KAZZAWARDS2024
• 540610 Tweets
SAROCHA REBECCA IN KAZZ
• 456964 Tweets
NORAWIT WITH PAUL SMITH
• 126997 Tweets
Varane
• 37744 Tweets
インプレゾンビ
• 33691 Tweets
魔法少女
• 25290 Tweets
B1NI TOPS SPOTIFY
• 24610 Tweets
FOURTH x KAZZ 2024🥳
• 23554 Tweets
#Destiny
• 18211 Tweets
#うたコン
• 16437 Tweets
名誉毀損
• 14118 Tweets
モグコレ
• 13166 Tweets
#仰天ニュース
• 12828 Tweets
風の行方
• 11763 Tweets
いなば食品
• 10216 Tweets
Pinned Tweet
📥You have 1⃣ new mail!
to: dearest VCs, angels, & assorted kindred souls.
from: me,
@andrewparkk
,
@SohamGovande
, &
@AnanthVeluvali
—————————————————————————
hi! we're 4 friends building in NYC this summer and we could use your help!
8
2
39
Go to Google Scholar and look up ‘As an AI language model” -“ChatGPT”’
536
8K
74K
airpods are a drug
ive noticed many young people (including me) wont do any mundane task (walking to class, laundry, eating) without an airpod in playing music
im a bit concerned abt how our brains are now used to constant stimulation and the threshold for boring has decreased
465
786
11K
A 🧵 of peer reviewed published scientific research where the authors left out a key coauthor 😉:
“As an AI language model…”
@MicrobiomDigest
118
1K
8K
Billionaires and regular people use the same phone. What are other things that don’t scale with wealth?
1K
223
5K
GPT-4 had 1.76 trillion parameters.
GPT-2 will have 100 quadrillion parameters.
Spread the word.
144
182
3K
i never believe recorded demos so I reached out to the
@cognition_labs
team for early access to try for myself and got it!
will be sharing my unfiltered opinions on
#devin
here.
🧵🧵 1/n
Today we're excited to introduce Devin, the first AI software engineer.
Devin is the new state-of-the-art on the SWE-Bench coding benchmark, has successfully passed practical engineering interviews from leading AI companies, and has even completed real jobs on Upwork.
Devin is…
5K
11K
46K
55
381
3K
Bridgit from good luck charlie working on a space tech startup in palo alto was not in my cards for 2023…
Also she’s moved on from MIT and is now at Harvard Law
35
124
3K
A thread:
A 🧵 of peer reviewed published scientific research where the authors left out a key coauthor 😉:
“As an AI language model…”
@MicrobiomDigest
118
1K
8K
9
84
3K
🧵 Thread of awesome AI projects by
@Stanford
students!
definitely gonna miss some projects so comment if i missed something!
(pic generated by stable diffusion)
62
494
3K
here's your DEEP DIVE into
@grok
's architecture!
I just went through the , for this 314B open source behemoth with *no strings attached*.
👇🧵
55
318
2K
reply ANY pic of the outside around you! 🌎🔎
i'll use a geospatial vision LLM to predict your location within twenty miles. 🕵️♂️
no landmarks needed. random trees/scenery works!
putting
@georainbolt
out of work
@geoguessr
554
158
2K
What if you could talk to the Bible?
Try now:
Describe your situation or ask a question! Cites specific verses from the Bible. Works in English and Spanish.
Made with
@OpenAI
and domain via
@Porkbun
!
#biblegpt
#chatgpt
#gpt3
it would only cost $4 to get vector embeddings for the entire Bible (~750,000 words at 0.0005 cents each)
it's wild that we've come so far and driven costs that low.
49
45
837
152
233
2K
Someone just dropped a dataset of 15 trillion tokens (as many as were used to train Llama 3)!!!
Download this now before it gets taken down for “copyright reasons”
Breakdown in thread 🧵 👇👇
32
143
2K
HOLY SH*T
@grok
IS 314 BILLION PARAMETERS
Mixture of 8 Experts, not RLHFd/moralized
THIS IS HUGE
28
137
2K
uh.... gpt2-chatbot just solved an International Math Olympiad (IMO) problem in one-shot
the IMO is insanely hard. only the FOUR best math students in the USA get to compete
prompt + its thoughts 🧵
🧵megathread of speculations on "gpt2-chatbot": tuned for agentic capabilities?
some of my thoughts, some from reddit, some from other tweeters
my early impression is 👇
40
83
543
73
172
1K
🔔the guy who invented the LSTM just dropped a new LLM architecture! (Sepp Hochreiter)
Major component is a new parallelizable LSTM.
⚠️one of the major weaknesses of prior LSTMs was the sequential nature (can't be done at once)
Everything we know about the XLSTM: 👇👇🧵
21
202
1K
Google’s Bard was giving out free WORKING microsoft office keys 😂
40
66
890
Deven is currently coding a Twitter clone
i never believe recorded demos so I reached out to the
@cognition_labs
team for early access to try for myself and got it!
will be sharing my unfiltered opinions on
#devin
here.
🧵🧵 1/n
55
381
3K
32
55
871
AI is so powerful I can predict someone’s location to the street in India from a single picture.
81
53
861
Thoughts on this tech company tier list (based on how nice it is to work there)
104
51
855
it would only cost $4 to get vector embeddings for the entire Bible (~750,000 words at 0.0005 cents each)
it's wild that we've come so far and driven costs that low.
49
45
837
Devin is now...training a Transformer LLM
i asked it to make its own version of
@karpathy
nanogpt
i never believe recorded demos so I reached out to the
@cognition_labs
team for early access to try for myself and got it!
will be sharing my unfiltered opinions on
#devin
here.
🧵🧵 1/n
55
381
3K
34
71
833
17
12
772
Is it just me or Quora has become unusable lately? So many ads and “related posts”. Used to be such a great resource along with Reddit
71
21
743
SATIRE SATIRE SATIRE THERES NOT ENOUGH COMPUTE ON THE PLANET FOR 100,000,000,000,000,000 parameters
15
4
751
Devin successfully gets llama.cpp running in a Docker container
i never believe recorded demos so I reached out to the
@cognition_labs
team for early access to try for myself and got it!
will be sharing my unfiltered opinions on
#devin
here.
🧵🧵 1/n
55
381
3K
24
46
716
is there a way to invest in people?
like here's $1,000 to buy 0.01% of your future net worth
a couple people i would love to do this for
191
27
665
CAN WE TALK ABOUT HOW BRIDGIT MENDLER FROM GOOD LUCK CHARLIE
USC undergrad, MIT masters and PhD, HARVARD LAW???
Oh And she’s working on a Stealth startup here in palo alto
15
33
620
PSA to not believe every AI thing you read online!
My “100 quadrillion” parameters pic is spreading around the internet
🛜
Most of big AI news accounts have no idea what they are talking about
67
39
611
gpt2-chatbot RETURNS! it's now TWO similarly performing models.
i've been testing them.
everything i can tell you 👇🧵
#gpt2
gpt2-chatbot was just turned OFFLINE
I was just using it half an hour ago!
@shaunralston
for the find
#gpt2
@openai
22
13
237
17
60
585
ChatGPT cites a fake article by authors X and Y (Year). Don't need GPTZero to catch that
2
27
544
I won 2nd 🏆(as a solo hacker!) at the
@XDevelopers
hackathon today!
Got to demo Flix to
@elonmusk
🙂
Thanks to the
@X
team for organizing this
68
22
554
🧵 Fish supply has dropped by 50% in 50 years...
That's why we built ShipSense! 🛳️
We coded a GIS portal to monitor overfishing hotspots 🔥 and built an AI vision model to detect ships from space 🌌
We also found a MASSIVE use case for
#StableDiffusion
👀
👇
#treehacks
#ai
40
69
538
@arithmoquine
Let me get you: quote retweet my original tweet
reply ANY pic of the outside around you! 🌎🔎
i'll use a geospatial vision LLM to predict your location within twenty miles. 🕵️♂️
no landmarks needed. random trees/scenery works!
putting
@georainbolt
out of work
@geoguessr
554
158
2K
3
10
545
🧵megathread of speculations on "gpt2-chatbot": tuned for agentic capabilities?
some of my thoughts, some from reddit, some from other tweeters
my early impression is 👇
40
83
543
This one is published on “” so who knows
Peer reviewed in the journal of internet banking and commerce
Stock price prediction is so tired 😴
10
29
496
If you want low rent, get out of the Bay Area.
Almost everywhere else on the planet is better for that. I'm not kidding at all. You'll be shocked by the stark differences in behavior of landlords in places where apartments are plentiful versus their behavior within artillery…
25
16
497
“Mr. Altman, this Chat PGT technology ahem predicts future tokens with a high degree of accuracy, as you said.
Could the Chat CPG be used by Russia to predict future winning lottery ticket numbers? And if so, what will you do about it?”
14
20
488
Tired of writing cover letters?
Cover Letter AI is a free site where you upload your resume PDF and a job description (JD). It uses
@OpenAI
GPT3 to analyze your resume and write a letter that highlights info relevant to the JD.
20
57
487
This shape has infinite surface area but finite volume. So you can fill it but you can never paint it.
40
44
489
CS majors learn computer SCIENCE which != software engineering/building
38
34
477
bridgit mendler is THE ONLY person who can rival that korean american Navy Seal Doctor Astronaut guy, and it's not close
she's only 30 so she has time...
Harvard JD
MIT PhD and Masters
USC BA
Space Tech Startup
Disney Channel Lead Star
@bridgitmendler
CAN WE TALK ABOUT HOW BRIDGIT MENDLER FROM GOOD LUCK CHARLIE
USC undergrad, MIT masters and PhD, HARVARD LAW???
Oh And she’s working on a Stealth startup here in palo alto
15
33
620
16
49
463
Grok is gigantic. Biggest open source LLM ever.
2x the size of GPT-3.5
Interesting that
@elonmusk
didn't share eval stats but that's kinda cracked bc evals are so broken anyways
here's your DEEP DIVE into
@grok
's architecture!
I just went through the , for this 314B open source behemoth with *no strings attached*.
👇🧵
55
318
2K
15
53
462
PART TWO of
#PaperGate
!
That
#ChatGPT
button "Regenerate response" often gets pasted into
#Scientific
Papers, via
@gcabanac
!!
A 🧵 of
#peerreviewed
scientific publications from reputable publishers like
@sciencedirect
@IEEEorg
@ElsevierConnect
First up, a paper on solar…
Go to Google Scholar and look up ‘As an AI language model” -“ChatGPT”’
536
8K
74K
18
112
456
🚀 Therapy costs $100+/hr. We’re fixing that.
@andrewparkk
and I are building
@meet_lily
, a voice-enabled AI therapist.
Lily sounds just like a human, is customized to each user, and responds in real-time (ultra-low latency).
Link + why we’re super excited about
#AI
therapy…
66
53
450
Breaking:
@Spotify
x
@OpenAI
Another massive partnership this week for
@OpenAI
. We saw
@BainAlerts
join hands with them a few days ago.
"The new feature is “a generative, expressive and realistic AI DJ that delivers a personalized lineup of music and …
15
43
452
best way to learn electrical engineering on my own? didn't get the chance
90
13
447
i asked the creator of the AI (
@heinenbros
) if i could publicly share the model, and he said yes (I was worried about overloading the server)
There's too many pictures for me to do them all, so go to the comments for the link to the LIVE DEMO for the AI👇👇👇👇🔗
reply ANY pic of the outside around you! 🌎🔎
i'll use a geospatial vision LLM to predict your location within twenty miles. 🕵️♂️
no landmarks needed. random trees/scenery works!
putting
@georainbolt
out of work
@geoguessr
554
158
2K
90
46
446
I made a Colab that turns a Github repo into 1 long but well-formatted prompt! Super useful for Long Context like
#Claude
Opus.
+ a map to show your LLM how everything is organized!
I used it to turn
@nomic_ai
's entire Github into one prompt, to ask questions.
Enjoy! Link 🧵
Does anyone have a simple tool that takes Github repo link and returns a well formatted string of all the code files? Ignoring non-code files
Would be very useful!
1
1
14
14
53
440
two of the greatest privileges you can have are:
US citizen
Fluent English
26
18
436
29
48
409
guess we got Google to take down these articles? moves very fast!
2
10
402
Actually insane: this person applied to 1,300 SWE internships
39
18
394
finetuned a stable diffusion model on myself today!
might make a beginners tutorial if people are interested
33
5
384
Someone alleges this about rabbit r1 and claims to have the source code 👀
46
28
387
synthetic GPT-3.5 data works amazing for fine-tuning smaller models!
generated a 13,000 Q&A dataset on the Aztec Empire in ten minutes. answers were designed to be sarcastic, upper-case, concise, and have an [END] token.
#llama
-2-7b (base) was able to learn all these guidelines…
11
41
372
grok is Apache 2.0 unmodified
thanks elon
5
26
360
The search criteria I am using:
Go to Google Scholar and look up ‘As an AI language model” -“ChatGPT”’
536
8K
74K
1
17
348
GPT3 had 175 billion parameters. GPT4 will have 99,220 quadrillion parameters. Image is to scale.
17
23
345
using Devin
@cognition_labs
to design DNA primers for flu detection
i never believe recorded demos so I reached out to the
@cognition_labs
team for early access to try for myself and got it!
will be sharing my unfiltered opinions on
#devin
here.
🧵🧵 1/n
55
381
3K
12
26
336
if gpt2-chatbot is GPT-5, we just hit a wall in AI
so, here's to hoping it's NOT, and is just GPT-4.5 fine-tuned
41
19
341
surprised i'm saying this, but
@AnthropicAI
's Claude Opus *is* noticeably better than GPT-4
Didn't believe it at first bc evals are shite but... wow.a
12
8
335
we won the
@Meta
most innovative prize and
@Replit
first place prize!
so thankful for my team
#treehacks
#ai
#sustainability
🧵 Fish supply has dropped by 50% in 50 years...
That's why we built ShipSense! 🛳️
We coded a GIS portal to monitor overfishing hotspots 🔥 and built an AI vision model to detect ships from space 🌌
We also found a MASSIVE use case for
#StableDiffusion
👀
👇
#treehacks
#ai
40
69
538
25
13
323
probably starting new thread bc this is getting long but thoughts so far:
Pros:
Devin feels UI/UX first, not Gen AI first. IDK if this makes sense. The AI is a core component but it's the surrounding infra they built that is the star of the show
The product feels quite…
40
14
262
my project for the
@XDevelopers
challenge
Turn any tweet into an AI video in any language
TikTok but on
@X
Uses
@grok
@xai
@elonmusk
40
21
304
25
19
305
thoughts on
@openai
's new website (why it was down earlier)
openai dot com now greets you with this search bar that launches chatgpt
i hate how the next slide (right of screenshot) is slightly showing
🧵
30
15
295
you wouldn't last an hour in the asylum where they raised me
20
11
294
Being a SWE in Washington is crazy. California comp levels but NO income tax (keep 30%!)
26
7
289
woke up to 20k, this is a dream
never expected to randomly become a tech twitter account back when i was a freshman making random tweets
31
3
279