E2E testing for AI agents that interface with users over messaging apps or A2UI is much smoother and faster than typical E2E testing for standalone web apps. Takes the dreaded brittleness of React/HTML/CSS components out of the mix.
English
ivelin.eth 🛡️🤖
6.9K posts
ivelin.eth 🛡️🤖
@ivelini
#OpenSource contributor and technology entrepreneur | @founding | https://t.co/jVd9McmctD | https://t.co/CDkdPePwdq | prior Telestax, Mobicents, JBoss/ @RedHat
Austin, TX Katılım Mayıs 2009
2.2K Takip Edilen1.7K Takipçiler
It is a moot point to argue whether humanoids will be the only useful robot form factor in the future.
Tesla vehicles with FSD are the most obvious example that ground transportation has very useful robots at scale in several formats: 2 seater (cybercab), 5,6,7 (S3XY), SUV (CT), Robovan, Semi.
Star Ship (SpaceX+xAI) is about to become the most useful robot for space travel or maybe even long distance flights on Earth.
Humanoids may cover the biggest number of use cases, but we have yet to see how many niche and large scale use cases will warrant specialized robot hardware.
English
At the pace entrepreneurs are realizing they can run multiple startups at minimal cost with AI agents , the notion of laser focus on one C-corp or series LLC seems to be transforming to laser focus on one Series C-corp or Series LLC.
It has always been a dual standard with investors demanding from founders to focus on one company , one idea only to maximize their chance of returns.
Thus founders taking extreme risk to spend 10 years of their lives for high chance of no return and a tiny chance of outsized return.
While investors themselves diversify across many founders and their companies to seek 10x portfolio return.
Well now founders can diversify too across many ideas with little or no funding at all. As long as they know how to build and manage multiple AI agent teams.
Founders can potentially get best of both worlds. Diversification and an increased chance for outsize return in one of their startups. All at the same effort and cost with proper AI agent leverage.
@bestape
English
Local DGX Spark running Qwen 3.5 35B A3B has proven a good coding buddy for projects around 10 KLOC.
Reliable snappy responses. No random cloud latency , arbitrary price spikes and frequent downtimes.
English
It seems that MiniMax M2.7 is more attentive and diligent with coding than Grok 4.20 Reasoning on ~10 KLOC project .
Grok repeatedly hallucinates for basic file actions it says it took, but did not. Good planning stage but coding stage does not impress so far.
English
@elonmusk Hard to tell if its improvement over 4.1 on coding. Seems to get lost and go in circles on projects ~ 10 KLOC.
English
What are your initial impressions of Grok 4.20?
Major upgrades are still landing every week.
Testlabor@testerlabor
Grok 4.20 is now officially out of Beta. It's now on Auto, Fast, Expert & Heavy.
English
The Fallacy of Complex OpenClaw Setups: Why Non-Technical Founders Should Start Simple (and Actually Win)
The influencer hype promises hyper-cosmic bot factories and 24/7 agent armies building your entire company or companies. For non-technical founders I recommend a simpler way to win.
English
@FinFreedom414 It’s a trick question. Life goes from A to B to A as kids age out of the house.
English
Imagine you had to choose your life at age 40:
Option A:
Single. No kids.
$10M net worth.
Travel anywhere. Total freedom.
Quiet house. Quiet holidays.
Option B:
Married. 3 kids.
$1M net worth.
Drive a Toyota. Chaos every morning.
Loud house. Full dinner table.
Be honest, which life are you choosing?
English
@herbertong I hope so. The question is on what time horizon. Grok 3 was going to be the last catch up model.
English
@elonmusk @Rainmaker1973 Cool support of the Matrix simulation theory. However according to @grok this Planck cube limit is still a theory that has not been experimentally verified.
English
@Rainmaker1973 The universe is fundamentally integer.
There are a finite number of Planck cubes, which means a limited number of digits of pi (which can be thought of in integer form) to calculate volume.
And you cannot have a fraction of a quark or lepton, so … integer.
English
Did you know?
NASA only uses 15 digits of π for calculating interplanetary travel.
At 40 digits, you could calculate the circumference of a circle the size of the visible universe with an accuracy that'd fall off by less than the diameter of a hydrogen atom.
π Day 2026
English
@elonmusk @peterwildeford Fingers crossed. Grok 3 was going to be the last xAI catch up model if I remember well.
English
@peterwildeford xAI will catch up this year and then exceed them all by such a long distance in 3 years that you will need the James Webb telescope to see who is in second place
English
Based on the data I see, I think:
- Anthropic🇺🇸/Google🇺🇸/OpenAI🇺🇸 all ~tied
- Meta🇺🇸 / xAI🇺🇸 each ~7mo behind
- Moonshot🇨🇳/- Deepseek🇨🇳 / zAI 🇨🇳 / Alibaba🇨🇳each ~9mo behind
- Mistral🇫🇷 ~1.5 years behind
- No other companies competitive
Ethan Mollick@emollick
Both xAI and Meta seem to be falling behind, based on the Grok 4.2 benchmarks and this reporting. Frontier AI models are really a three way race at this point.
English
And Tesla may be the invisible undercurrent that sinks Apple’s chance to rise with the AI tide.
FSD hardware is as reliable and smooth as consumer life critical AI can possibly be.
And it is about to work on general purpose workloads when parked. At RTX 3090/DGX Spark / Mac Studio performance.
Right behind it , a wave of Optimus fleet running the same and next gen Tesla AI chips…
English
Unified memory architecture is a winner for edge AI. If Apple won’t wake up, NVIDIA will dominate.
DGX Spark has been a fantastic little box so far. Three months of no issues.
Beautiful design. No hang ups. No upgrade glitches. No unplanned reboots. No overheating.
Solid. Mac grade consumer experience so far.
Limitless@LimitlessFT
NVIDIA as about to announce their OpenClaw killer with their new platform 'NemoClaw', the open source AI agent platform. NVIDIA also just committed to spending $26B on opensource AI models over the next 3 years. This is either a genius play by Jensen... or a colossal mistake. We unpack this (and much more) in todays episode 👇
English
Tesla FSD HW4/AI4 dual chip with 16GB RAM compares to RTX 3090 for inference , according to @grok. That is a lot of firepower.
3090 beats the latest DGX Spark on FP16. I have both and verified this.
Having another 3090 in the driveway running nightly AI cron jobs is appealing.
English
@CharlieMcGill20 @Jason @openclaw @Tesla_Optimus That’s far from clear. Tesla has been gradually opening the APis and apps for cars. Why not for Optimus?!
English
Jason… respectfully: For someone who is so close to Elon personally, you really struggle to understand his business intent.
There’s is absolutely 0% chance Tesla lets you operate ANY software on Optimus. It will be Grok/ Optimus’ own software only!
Makes me wonder if you just say things to pretend not to have insider info ;)
English
so.... eventually.... I'm going to port my @OpenClaw in a real-world @Tesla_Optimus, and it will run all my jobs inside its local brain (and I can sell my Mac Studio)?!?!?
ALL IN! (so to speak!)
Elon Musk@elonmusk
Macrohard or Digital Optimus is a joint xAI-Tesla project, coming as part of Tesla’s investment agreement with xAI. Grok is the master conductor/navigator with deep understanding of the world to direct digital Optimus, which is processing and actioning the past 5 secs of real-time computer screen video and keyboard/mouse actions. Grok is like a much more advanced and sophisticated version of turn-by-turn navigation software. You can think of it as Digital Optimus AI being System 1 (instinctive part of the mind) and Grok being System 2. (thinking part of the mind). This will run very competitively on the super low cost Tesla AI4 ($650) paired with relatively frugal use of the much more expensive xAI Nvidia hardware. And it will be the only real-time smart AI system. This is a big deal. In principle, it is capable of emulating the function of entire companies. That is why the program is called MACROHARD, a funny reference to Microsoft. No other company can yet do this.
English
@abhidinesan @Jason @openclaw @Tesla_Optimus For sure. It’s quite obvious by now. There is still no openclaw config way to use xAI grok with its native web and X search tools. Openclaw keeps forcing client side tools. Issue and PR have been open over a month. Building a custom skill is the common workaround.
English
@Jason @openclaw @Tesla_Optimus Open source for now… there’s a non-zero probability that OpenAI does its signature move and takes it private.
Also feel that they are probably favouring a super tight vertical integration that’s optimized for Tesla Silicon.
Do hope your right tho, would be awesome to see
English
Point of no return. @coinbase , @RobinhoodApp , @krakenfx and @Nasdaq with many other pioneers who did not quit too early.
Mark@markchadwickx
I don't think people understand how insane this actually is... Nasdaq just partnered with Kraken to enable 24/7 trading of tokenized stocks. The craziest part is it'll run on Kraken - not through traditional brokers like Schwab or TD. Last week the Fed, FDIC, and OCC confirmed tokenized securities receive the same capital treatment as traditional securities. Kraken also just gained access to Federal Reserve payment rails, putting a crypto firm inside the same core settlement infrastructure used by major banks. Think about what that means. A crypto exchange is now becoming the platform where tokenized equities trade globally, 24/7, 365 days a year. This is how trillions move on-chain. We're in the calm before the storm. Mass adoption is coming and Mass rotation of capital is coming with it.
English