Sam
@0xdeadlift
Followers
567
Following
719
Media
250
Statuses
1,436
Indie Game Dev / AI Whisperer ◦ Ex @coinbase Infra◦ Building LLM-powered worlds w/ ◦ I like having fun online
San Francisco, CA
Joined July 2014
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Puerto Rico
• 559593 Tweets
Cowboys
• 60569 Tweets
紙の保険証
• 58070 Tweets
せなけいこさん
• 39212 Tweets
Mustafa Kemal Atatürk
• 34132 Tweets
ねないこだれだ
• 32352 Tweets
不在者投票525人分
• 29071 Tweets
大阪・豊中市
• 28083 Tweets
車の中の段ボール箱
• 27506 Tweets
Efendiler
• 25325 Tweets
$ICE
• 23425 Tweets
Duterte
• 21439 Tweets
Happy New Week
• 20465 Tweets
C-295
• 19768 Tweets
#リリリエラ
• 15187 Tweets
KabinetPRABOWO GIBRANkompak
• 14557 Tweets
Langsung GASSS Kerja CEPATTT
• 13838 Tweets
コンパス
• 12598 Tweets
Bato
• 11779 Tweets
Pinned Tweet
Don't have a big October Xenoworlds update this month but I did want to share a bit of what I've been working on – mostly I've been reworking the initial stages of Xenoworlds based on feedback received during playtesting
First, as cool as the midjourney placeholder faction
0
0
7
@erla221942
guy was such a badass and came so close to making it – he was only a few floors away from a machine floor open to the air where he could've climbed back inside and headed down the stairwell
6
10
3K
@MorrowindGOTY
Ngl I immediately bought Alpha Centauri after seeing that post last night. Wonder how many sales it’s gonna push for EA
1
0
200
@bcmerchant
I don’t think this comparison is saying what you think it is….they look similar because openAIs video is an accurate representation of what a shot of a Victoria crowned pigeon looks like
Are they supposed to make it the wrong color or shape so no one can claim it’s a “copy”?
4
0
143
@datepsych
This shit is super funny to me bc as a natty lifter it’s actually quite the compliment to have someone accuse you of juicing
4
0
73
@TheEdmvnd_
Ah yeah I realize that. Was just remarking on how it’s doubly confusing for us but didn’t want to delve too deeply into it 😁
4
1
64
@buildren
@bcmerchant
Yes I am. In fact, it looks like all videos of Victoria crowned pigeons because that’s what Victoria crowned pigeons look like
2
0
62
@postjawline
Such a vibe, imagine climbing around there in a space suit, knowing a single wrong move in the tenuous gravity could send you spiraling out into the black
2
0
54
@somewheresy
This must be that man-made horror beyond our comprehension they’re always taking about
0
1
43
@charlosrossi
@Jay_Andy12
@erla221942
that’s it, the floor was open to the air to help vent the machinery inside
2
0
36
@TheEdmvnd_
This joke is super confusing as an American bc while we do use unlock all the time, we’d still definitely be staring in disbelief
In America cops can’t just demand you unlock your phone without a warrant
3
0
31
@MarioBrothBlog
People keep making political compass jokes but I just wanna know
Is this loss?
0
0
30
@Paracelsus1092
Nah, Yuval is right here, at least for the first world. In 20 years we’ll have AGI and it’s really hard to predict what humans will be doing once “human work” is no longer the core piece of the economy — this has obviously never been the case in human history
10
0
29
The fact that we haven’t heard a peep from OpenAI on the 1 year GPT4 anniversary is definitely weird — makes me think we might get something nice tomorrow
no Blog ?
No tweet from leadership about the lead up to last year ?
No Dalle birthday cake from
@ChatGPTapp
?
No Sora video to mark the day ?
GPT4 having his birthday ignored like Harry Potter.
Patient Jimmy it is I guess.
0
40
563
5
0
28
Rough week for the leakers with no new OpenAI model...gotta give credit to
@futuristflower
for spotting this one ahead of time though
With OpenAI, Figure 01 can now have full conversations with people
-OpenAI models provide high-level visual and language intelligence
-Figure neural networks deliver fast, low-level, dexterous robot actions
Everything in this video is a neural network:
1K
5K
17K
0
0
24
@BenjaminDEKR
Probably a function of everyone’s expectations for gpt-5 but it also definitely feels weird for them to be pushing an incremental update from all the big accounts with no benchmarks
2
0
22
Perfect JSON structure??? 40x the speed of GPT4???
Introducing Inversion, our family of structured LLMs.
Our first generation models excel in structured tasks, offering unmatched speed, latency, reliability, and efficiency, with the most comprehensive typed JSON output support available anywhere.
62
135
951
1
4
23
April Xenoworlds Update
This month I started working on a new mode I'm calling "Worldshaper". Unlike previous versions of Xenoworlds which generated the entire world at the start, the gameplay revolves around building up the world as you play.
You start by selecting some details
2
0
20
cookin up some cool new stuff right now – will be grinding this week since I'd really like to get a playable tech demo out by end of month 🤞
6
1
21
@Aristos_Revenge
This thread is full of language model misconceptions but I think the biggest issue is the idea that some other company will come up with a less restrictive AI and beat chatgpt…you can use gpt3 directly right now, the less restrictive AI is here and publically accessible already
3
2
20
Inference speed on
@GroqInc
is insane, just blows everyone else out of the water
When gpt4 came out, everything generated by gpt3 immediately started feeling primitive in comparison
Now everything producing less tokens/s than groq feels viscerally slow
Try now
Hyperfast LLM running on custom built GPUs
Answers in miliseconds, not seconds
How? 🤯
169
205
2K
3
5
18
@eshear
I have a very vivid memory of looking at my ATM balance one day and realizing “this is literally just score points I’ve managed to accumulate”.
Haven’t been able to shake that frame since
1
0
17
@fchollet
That's great to hear! My shoddily over-engineered code will allow me to fit in anywhere 🙌
0
0
16
@mattyglesias
Oh, so now we're just moving on to blaming the Jews and their weather machines instead?
1
0
17
@wrathofgnon
I went there last year! The hotel I stayed at had a comically large map of the area, probably to distract from how big the lobby was for so few guests
The attached town nearby seemed basically dead
0
0
15
January Xenoworlds Update: Made a ton of progress this month, with solid development on game visuals as well as movement on a bunch of core gameplay systems
First off, worlds have tiles now! Map regions have existed for a while now, but they were based on a ~122 square grid,
5
0
14
Unfortunately, what the people want is GPT-5, and what the people NEED is dirt cheap/fast haiku beating 3.5-turbo-turbo, not another 4 version
2
0
14
Actually starting to think OpenAI’s coyness around model benchmarks may mean the new model is state of the art
Wouldn’t be surprised to see them retake the lmsys leaderboard top spot when it next updates
4
0
12
biggest impediment to productivity right now is that Slice & Dice is an insanely well designed game and it’s taking forever to optimize the fun out of it
4
0
11
@SofiaErnestina
Hi Sofia, I can explain!
In 7300BC the Finno-Korean hyperwar reached its zenith, culminating in a thermonuclear war that ended the ice age
Now, Earth has temperatures much higher than previously, which DETERS the hollow earth powers from invading the surface
Hope this helps!
1
0
10
My latest project, Xenoworlds is getting pretty close to alpha, so it's time to start talking about it a bit. Xenoworlds is an "Anthropological Exploration" game – I've built a literal world building system that creates entire worlds to explore using AI
#AIgamedev
#aiart
2
3
12
@TeamMGG
Tbh it probably wasn't a smart move to get rid of your most popular and charismatic player but I'm glad to see proof in the open you guys have no principles and easily bow to pressure
1
0
9
February Xenoworlds Update: A lot of the work for this update was refactoring behind the scenes, so there's a little less showy stuff this month versus January.
Operatives: The biggest piece of development work this month was in creating "Operatives", the main piece of
2
1
10
Can you imagine being stuck on obsolete ribeye v1 when super ribeye 3000 is available??? I’ll bet lab grown tastes better than the real thing within ten years
Global elites want to control our behavior and push a diet of petri dish meat and bugs on Americans.
Florida is saying no. I was proud to sign SB 1084 to keep lab grown meat out of Florida and prioritize our farmers and ranchers over the agenda of elites and the World Economic
1K
2K
14K
2
0
8
this is the inverse cramer indicator of AI right? GPT5 release must be imminent
2
0
9
@npceo_
@DeveloperHarris
This is huge and something I also think about all the time – building context and structure for LLM games isn't just a nicety, it's a requirement.
LLM output is like water in that it will fill whatever shape you provide for it. If you don't provide a shape at all, it'll go in
0
0
10
Got everything swapped over to llama 3 and it's a huge improvement in terms of creative writing – check out llama 3 here relating the culture of the Ra'kreshi explorers to the behavior of the animals they just discovered
gpt-3.5 could never
0
0
9
May Xenoworlds Update
Update is a little later than usual this time since I took a little eclipse vacation last month and spent a bunch of time on game design – still managed to get a bunch of new systems built out, so let's talk about it!
Biggest change for this update was
4
1
9
@pronounced_kyle
2030 honestly can’t come soon enough!
Hopefully we’ll get positive results so we can send over one of these bad boys
0
0
7
@postjawline
This article is nice wishful thinking but whenever I talk to normies about AI the number one emotion is “fear”
0
0
9
ngl I am definitely enjoying my life of coding up video games every day. 13-year old Sam would be so psyched
0
0
7
The hardest part of building a game hasn’t been the ui, or the story, or even the code — it’s 100% been the game design
0
0
6
81 degrees in SF tomorrow and I couldn’t be more hype. This is as good as it gets here baby, our one yearly day of summer
2
0
7
Can’t say I’m really scared of Devin
As an infrastructure dev half your job is automating away whatever you did last year — and every year there’s still infinitely more things to automate
AI devs just means you can run the cycle even faster!
0
0
7
It's insane how much more I enjoy coding now than when I quit my job
Felt so burnt out I was considering taking the LSAT and getting a law degree instead!
But then I installed copilot, started building for myself, and it alllll came back
1
0
7
Summertime and the living is easy
81 degrees in SF tomorrow and I couldn’t be more hype. This is as good as it gets here baby, our one yearly day of summer
2
0
7
0
0
8
swore to myself in 2017 there was no way in hell I’d miss totality for the next one…and here we are
0
1
8
Finally internalized this lesson a few weeks ago and it totally changed the way I structure my api calls. Inference cost seems likely to trend to ~zero over time
I think this will also probably apply to generation speed, as we’ve already seen with improvements from
@GroqInc
How many ideas have you killed early because you thought they were “too expensive with LLMs”? Costs have come down 100x already - plan for 100-1000x more. Show the world your crazy ideas! 💖
2
2
29
2
0
6
There was already a lot of anti-AI sentiment in the art world but I feel like the Sora release might be the point where things start to spill over into the wider cultural arena, unfortunately
Politicization of AI is inevitable as capabilities increase
3
1
6
gym I’ve been going to for half a decade is closing this week. but not before I could get in one last PR
2
0
7
wtf are "bytearray" people? another weird llama3 quirk is it sneaks in shit like this every once in a while
2
0
7
Beautiful day in SF today! Time to shut my blinds and be a little code gremlin all day
0
0
6
May Xenoworlds post is taking a little longer – I've been refactoring a few systems and need to reintegrate some of the stuff I planned on showing off.
Bigger update is coming soon, but here's a preview of the new civ creation "concept cloud" style I've been playing around with
0
1
7
It’s cyberpunk as hell we still don’t know who satoshi is and I hope we never find out
Think about how big of a secret Satoshi Nakamoto’s identity is 🤯
This dude/dudette/group created a $1T asset class and their identity has never been revealed
If a secret that big can be kept, what other similar scale secrets exist?
Are there any that are 10x bigger? 100x?
176
53
1K
0
0
1
Entering my unhinged posting era, we out here posting whatever the fuck comes to mind
Hinges 👎👎👎
3
0
6
NYE Xenoworlds update: The real time system is fully implemented and working – events now stream in regularly and you can change the speed at which time passes on your world. I added a timeline to the main screen to show when upcoming events will occur
#indiegamedev
#aigamedev
1
1
7
I don’t think there’s a single time I’ve been reminded of supersize me that didn’t end with me getting McDonald’s shortly after
0
0
5
lmaooooo turns out Claude is actually the best model out there, it’s just that Anthropic’s alignment methods have real costs
We just launch a new Category "Exclude Refusal" based on a keyword filter to identify whether model refuses to answer user question (e.g. "sorry, I cannot answer...")
We filtered out ~7% (40K) votes. Result:
1. All Claude models go up, despite still 1>2🤔
2. Some open models
7
19
109
0
0
5
Been jonesing hard to integrate llama 3 into Xenoworlds and should get there this week — I am very excited
1
0
5
Had a fascinating conversation with a relative in show business about AI and the movie industry yesterday and got to see behind the scenes on the SAG strike
He said there was a big split amongst the actors about whether the focus should be protections against studio AI use or
0
1
6
lmao, this image is borderline body horror. How did this ad possibly pass through quality control?
3
0
6
got llama3 working on
@GroqInc
but their insanely low default rate limits mean it's basically non-functional even when it's just me testing things out
2
0
5
Would like to extend a warm welcome to all the new members of the Drake Haters fan club. We've been hating on that degrassi-ass buster for over twenty years now, thank u for joining and supporting our righteous cause
2
1
5
have been kinda bored with twitter for the first time since I entered my poasting era. feels like my head was bursting w tweets a few months ago but now they’re kinda hard to come by
1
0
6
apparently this works for rocket ships too??
0
0
5
@hkarthik
I used to work for Coinbase but quit a few months ago — the reason you don’t see SF employees layed off is because it’s not an hq and there is almost no one left there.
There were probably <40 people per day coming in when I left and zero execs setting foot inside the office
0
0
6
Sooo where can I go to eat the lab grown meat? I’m ready to taste the future
0
0
6
this is definitely a win for america – this was already de facto true in california and its been a huge boon for silicon valley
BREAKING: The FTC just banned non-compete agreements, per MorePerfectUnion.
The Federal Trade Commission has issued a final rule making it illegal for bosses to make workers sign noncompetes in any scenario, and voiding nearly all existing noncompetes.
817
6K
39K
0
0
5
I’m tired of Uber eats, when waymo eats? Robot delivery adds to the flavor
1
1
5
Not sure I'll ever forget the insane 500+ comment google doc discussion that ensued when Coinbase eng leadership tried to mandate using Mongo instead of Postgres
Interesting to see this cycle happen again and again.
Most startups start out as "use whatever tools you want - we're a startup, after all!"
Fast forward several years when the company uses N languages/frameworks. There is now a valid business need for (some) consolidation.
32
22
286
0
0
4
Checked out llama 3 last night and it's kind of cracked – GPT4 quality writing at prices a fraction of GPT3.5
Feels like zuck knocked it outta the park with this one, unless openAI drops a new model soon I really can't see why a reason to choose them OR anthropic over llama
0
0
4
good morning from beautiful Flippin Arkansas
population 1,345
most well known for the 1990s Bill Clinton “Whitewater” scandal
1
0
4
@GhostOfNolte
@Aristos_Revenge
Second the idea that Sam Altman and openAI are trying to demand the AI act like an extreme neolib. This is kinda an accidental artifact of the training. Sam's last interview actually mentions he believes models should give the output user's want rather than what a company wants
4
0
5
This is the way
0
0
5
The highly rational scientists of the Particula fear nothing more than the mutants living deep beneath the streets of Excavatori.
But their much-vaunted rationality fails to predict the successful joint society should the Free Mutants take power
#aigamedev
#aiart
#xenoworlds
0
0
5