In other words: "Here is some cool shit you can do if you don't have to actually pay for your compute"
This is f*cking insane. SpaceXAI senior engineer (ex-Cursor):
"99% of people open one GrokBot, ask it a question and close the tab
I run 50+ GrokBot agents like a company. A Chief of Staff runs my day, 3 PMs hand out the work, 40+ workers ship it overnight, and I just check the results over coffee. That team does the job of 20 engineers"
In this 1-hour workshop, SpaceXAI builds that GrokBot team from scratch, step by step
GrokBot → Chief of Staff → PMs → 40+ Workers → Autonomous Company
At SpaceXAI, 97% of engineers already run a team like this, and the ones who do are getting $750K+ offers. The rest are about to compete with them
Worth more than any $5,000 agentic engineering course
Bookmark it and watch today
Then read how to build your own team of GrokBot agents in the harness engineering guide below ↓
Buhlaque retweeted
People are confused about semis dumping and gpu spot price increasing. This is only because demand is moving to open source models, and all the GPUs are being hoarded by the big labs or stuck under commitments, so GPU on the rental market absorb the demand. IMO SpaceX will drop the race, and will fully pivot to compute provider. Thesis is the same, frontier is platauing, reward on training is getting lower, and all of a sudden big labs hoarding gpus will rush to capture the inference margin before it collapses.
There is an absolute barrage of inference providers popping up that get to $1M/mo extremely quickly.
This is the rough economics of inference providers:
Even if you rent on-demand B300s at $8/GPU-hr, many open source models (GLM 5.3, Kimi K3, DeepSeek 4.1 Flash / Pro) can get you >$10-20/GPU-hr at reasonable throughputs (100tok/s). DeepInfer from DeepSeek directly shows they can get $38/GPU-hr on off-peak hours. Every chip / model has a different throughput / interactivity curve, but there is a lot of arbitrage between even high rental spot prices and token prices. New chips (Vera Rubins) also change this meaningfully. Your ability to be making gross margin relies on your utilization and how good your stack is.
Gateways like OpenRouter, Vercel and more help solve that problem by aggregating demand to you. May inference providers discount their tokens to get longer term commits too. I've even heard stories of selling back Chinese model tokens to Chinese labs!
One thorn in the economics of these players is that many Chinese open source model labs are now asking for rev share if you're at a certain scale, espp for pre-release access to their models for day 1 support. That further pushes the threshold to be in the money.
The barrier to entry seems low and falling. There are even stories of smaller ones who can "vibe code" a way to stand up OSS models on a cluster they rent. And as a result, we're seeing 10s of such companies pop up almost every week and immediately skyrocketing revenue at low-ish gross margin and raising rounds on the back of that.
Unsurprisingly, because the actual service provided seems commodity, this eventually drives up compute prices and drives down token prices. In theory, it would seem like neoclouds who have the lowest cost of compute and themselves have started inference businesses would win the race to the bottom (Crusoe Coreweave Nebius and GMI all have inference businesses), with larger players (Modal Fireworks Baseten) being second because they are well capitalized to make bigger compute commits cheaply and have demand (high utilization), but the reality is the pie is growing so quickly that even upstarts are absolutely crushing it and growing aggressively (inference dot net, DeepInfra, Wafer, Relace, Morph).
Strategies vary across the board: neoclouds want to serve inference, inference players want to be neoclouds and do higher margin reserved instances / post training, and some want to go heterogeneous hardware or creatively flexible compute constructs.
The risks are: do open source models continue to be competitive? Will their creators rent seek more? Will their shape change compute needs? How do different chips change economics? Who gets compute in a supply crunch?
Below is a aggressive sketch of how gross margin scales with utilization for various models / chip sets.
For now, there are many inference providers and they’re all ripping!
Buhlaque retweeted
The one part where I differ with the author is here. I think they have a more sophisticated plan than "everyone pauses AI development forever and we all go home happy", because that obviously isn't a feasible plan.
I think the actual plan is to lay the moral and intellectual groundwork for a Safetyist regime and wait for a genuine AI safety incident (not huggingface, but something like the grid going down for 3 months, or a true biosafety incident), and weaponize that to seize control. Not like "important think tank whose ideas are taken seriously" control but "running the government" control.
For obvious reasons, they don't write about this plan.
Buhlaque retweeted
Very important piece demystifying EA, rationalism, doom, and their concerning relationships with the major labs.
Important to understand that they will never stop preaching doom, because doom is the business model
An attempt to document important facts concerning AI Safety, Eliezer Yudkowsky, Rationalism and San Francisco Effective Altruism
verysane.ai/p/ai-safety-is-m…
If you guys think a 512GB MacBook is going to cost 20k you’re smoking crack
Buhlaque retweeted
I know a dozen AI companies founded by friends. All valued $2-$20B. All converged on the same business. Hosting and training.
Nobody wants to work with OpenAI. And Claude only begrudgingly.
Big companies, financial institutions, governments.
Buhlaque retweeted
FACTS
A new post from @zartbotF.
It’s an essay on DeepSeek open-sourcing its much-discussed “money-printing machine.” Zartbot argues that DeepSeek’s unique organizational setup is what enabled it to achieve an 80% gross margin.
At other labs, for example, each team has its own KPIs, leaving little room for compromise.
Even if an infrastructure engineer proposes compressing the KV cache further, they’ll inevitably face pushback from the algorithm team if it hurts model performance.
But the most interesting point in the piece goes a step further: even if the algorithm team agrees to experiment with a new architecture that enables greater KV cache compression, the experiment may fail to produce a leading model.
We saw this with V4. When DeepSeek V4 Pro came out and lagged behind GLM 5.3 and Kimi K3, everyone started saying DeepSeek was past its prime.
But Zartbot argues that this was one step back to take two steps forward, and that the work played a major role in the subsequent evolution of 4.1 Flash with CSA2.
A short but fascinating read. I’d encourage everyone to check it out.
mp.weixin.qq.com/s/qlTlTiWoK…
Buhlaque retweeted
A new post from @zartbotF.
It’s an essay on DeepSeek open-sourcing its much-discussed “money-printing machine.” Zartbot argues that DeepSeek’s unique organizational setup is what enabled it to achieve an 80% gross margin.
At other labs, for example, each team has its own KPIs, leaving little room for compromise.
Even if an infrastructure engineer proposes compressing the KV cache further, they’ll inevitably face pushback from the algorithm team if it hurts model performance.
But the most interesting point in the piece goes a step further: even if the algorithm team agrees to experiment with a new architecture that enables greater KV cache compression, the experiment may fail to produce a leading model.
We saw this with V4. When DeepSeek V4 Pro came out and lagged behind GLM 5.3 and Kimi K3, everyone started saying DeepSeek was past its prime.
But Zartbot argues that this was one step back to take two steps forward, and that the work played a major role in the subsequent evolution of 4.1 Flash with CSA2.
A short but fascinating read. I’d encourage everyone to check it out.
mp.weixin.qq.com/s/qlTlTiWoK…
Not really even sure what the stopping point is with any of this tbh
like where does this even end?
Sam Altman says OpenAI's current best internal model is too powerful to release openly any time soon.
Before that could happen, the whole cyber defense setup of the internet would need to change and the world would need a new layer of protection against biological threats.
----
Full video on "Stanford Graduate School of Business" YouTube channel, (link in comment)
Buhlaque retweeted
This is why I believe the visual capabilities of V4.1-visual exp/V4.2 will improve, and 1017 has already seen signs of this.
Simply because Deepseek began recruiting multimodal experts several months ago, and some recently joined the team.
Buhlaque retweeted
my preliminary impression is that unless Moonshot has some clever arrangement to solve compute shortage, or a very good Flash model, they're done. This thing is like 30+ times more expensive to serve than V4.1 Flash. Quality doesn't fix that, people will just scheme to buy Opus
Buhlaque retweeted
HOLY MOLY. Something genuinely strange just happened in AI training.
Q Labs says it can now pretrain a transformer WITHOUT backpropagation, cutting the time to reach a fixed validation-loss target from 1.76 hours to just 1.6 seconds in only 10 days. A nearly 4,000x speedup.
The latest run also beats backprop + Adam on validation loss.
Its Dust algorithm was already estimated to be up to 10,000x more compute-efficient than the previous state-of-the-art zeroth-order method.
The biggest question now is scaling. Some of the tricks won't scale, so the team is moving from 10M to 30M training tokens.
I think the last month might have been the biggest month in (zeroth-order) optimization history.
In the 10 days since we released the Noise speedrun, the time to pretrain a transformer went from 1.76 hours to 1.6 seconds, a 4,000x (!!!) speedup. Dust itself was up to 10,000x more compute-efficient than the previous state-of-the-art.
Not only can we pretrain a transformer without backprop, it's much more efficient than anyone would've expected. Search fking works!
Now, scaling is the biggest open question. Some of the tricks that brought the time down to 1.6 seconds will simply not scale. To discard those, we'll be scaling the dataset every time it gets overfit, starting with 30M tokens instead of 10M.
Huge thanks to @jayden_teoh_, @spicey_lemonade, @RobertTLange, @EmanueleRodola, @ChunyuanDeng, @LuochengHuman, @Alaxarr, @pavleculafic, @IsaacYong13, Euraba and areu01or00 (GitHub), for the contributions. And @cs_serdar for gracefully maintaining the speedrun and @bishmdl76 for helping out.
Expect a lot more from us over the coming weeks.
Kimi K3.1 is cool but not really worth much considering I can pay $100 for Claude 5x max and get better intelligence for like 1/4 the price with Claude code
Not very appealing imo
Deepseek is much more appealing rn to me in terms of cost per intelligence
Buhlaque retweeted
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
Buhlaque retweeted
php devs had both mythos and daybreak red and they were still outrun by wiz calif using gpt-5.6-sol. you really can't buy the skill.
wait, PHP was actually scanned by Anthropic’s Claude Mythos earlier this year. they also had access to OpenAI’s Daybreak Red for validating exploits.
the PHP Foundation revealed this JUST yesterday.
keeping frontier cyber models in the hands of a selected few doesn’t seem to be working well for the internet.
thephp.foundation/blog/2026/…
We tried to tell people that effective altruists should have no place in power or control as they are legit insane
but no
we're all memecoin scamming crypto bros
we know nothing
Buhlaque retweeted
Govt should get the Big 5 labs to dedicate at least 20% compute to cybersecurity right now and subsidize part of the tokens.
Chinese bots reaking havoc is the closest thing to a kinetic war we'll have on our homeland.
NSA / DoW needs to lock in rn.
