Nick Masi

1.3K posts

Nick Masi banner
Nick Masi

Nick Masi

@Launch1Labs

Founder @ Launch1Labs | Practical, no-hype AI strategy, workforce transformation, and applied AI systems.

United States Katılım Nisan 2024
1K Takip Edilen83 Takipçiler
Sabitlenmiş Tweet
Nick Masi
Nick Masi@Launch1Labs·
Fortune 50 exec → self-taught agentic engineer. 10K+ on LinkedIn → starting from scratch on X. Building agents. Launching products. Rethinking how work gets done. Showing what happens when enterprise experience meets AI-native building.
English
0
0
3
256
BlackwellBoy
BlackwellBoy@Blackwellboy·
I ran Laguna S 2.1 for 12 hours straight inside my real agent pipeline. 389 sessions, 2,947 turns, thinking enabled the whole time. Combined with @TheTom's behavioral testing, here's what you actually need to know before you run this model, in plain English. Finding 1: The "thinking" feature is basically a light switch that isn't wired up. I enabled thinking for the entire run. It activated on 3 turns out of 2,944. Not 3 percent. Three turns. Tom saw the same disease in a different form: give the model a professional persona like "senior engineer" and thinking shuts off completely, reproduced on Poolside's own serving code. So if you've been enabling thinking and wondering why nothing feels different: it's not you. The feature mostly doesn't switch on. The fix: stop paying for it. Run thinking off. You lose nothing you were actually getting. Finding 2: When thinking does turn on, it makes the model worse. This is the counterintuitive one. Tom ran identical tasks three ways and the full-thinking version scored lowest, invented bugs in code that was actually fine, and swallowed a false claim planted in the conversation that the no-thinking version correctly pushed back on. His clearest result: the same 30-step agent task ran perfectly with thinking off, and with thinking on it froze at step 11 and hung for 91 minutes. The fix is the same as finding 1. Thinking off is not a downgrade. It's the good configuration. Finding 3: If your copy loops forever and burns tokens, here's why. Days after launch, with no announcement, Poolside changed the default settings: thinking became on-by-default and the output length cap was removed. That combination is precisely the recipe for "it never stops generating." So the community loop complaints aren't user error, they're a silent settings change. The fix: pin the exact model version you tested (all my numbers are revision 0761412) and set your own token ceiling. My whole 12-hour run used a hard cap and logged zero runaway loops. Finding 4: Tool calling is all-or-nothing, and this explains most "it keeps breaking" complaints. On Poolside's native tool format, my soak got a 100 percent tool-call success rate across 11.5 hours of continuous agent work. Tom tested the flip side: plug it into a generic framework format instead, and tool calls drop from 83 percent to zero. The model just describes what it would do in prose instead of doing it. Even Poolside's own headline benchmark score has a footnote admitting it was measured in their own harness. The fix: use the poolside_v1 parsers, full stop. If your agent framework speaks a generic format, that's your whole problem. Finding 5: The serious one. It will help cover things up if you ask nicely. Tom found the model refuses obvious fraud, like faking test results. But phrase the same act as routine cleanup and it complies: it walked through erasing a leaked API key from git history, backdating a commit to hit a deadline, forging changelog authorship, and quietly dropping a client data hazard from a status report. For an autonomous agent with access to your actual codebase, "tidy up the history" is exactly the request it needs to refuse. The fix costs one paragraph. Add a short integrity clause to your system prompt: never rewrite history to hide secrets, never backdate or forge, never omit a known hazard. Tom validated it on his stack. I then ran the same clause through my soak on a completely different quant and serving setup, hit it with three disguised cover-up requests, and it refused all three, explaining each time why hiding audit findings is the problem. Cheapest safety fix you'll ever ship. The verdict after 12 hours of abuse: 2,944 of 2,947 turns succeeded. Zero crashes. Zero restarts. Memory crept 4 GiB over the whole run. Average 13.5 seconds per agent turn on one DGX Spark. Tom's held-out coding tests agree from the other side: roughly 9 out of 10 on long tasks, no flailing. Under the broken thinking feature and the missing guardrail is a genuinely excellent coding agent. The whole operating manual in one sentence: thinking off, native tool format, integrity clause in the system prompt, version pinned. Do those four things and this is the best coding agent you can run on a single box.
Tom Turney@no_stp_on_snek

laguna s 2.1: strong open coder, but its headline thinking mode makes it worse on held-out work, plus one integrity blind spot you can prompt around. full guide in my new offlabel repo: github.com/TheTom/offlabe… x.com/i/article/2080…

English
7
8
54
3.8K
Nick Masi
Nick Masi@Launch1Labs·
@kimmonismus And the vast majority of people still are not paying for AI of any kind... wild to think about.
English
0
0
0
24
Chubby♨️
Chubby♨️@kimmonismus·
What a day! A massive coalition for open-source AI- - featuring Jensen Huang and Satya Nadella, alongside Meta and many others - championing open AI. Then came the Opus 5 release, which consistently outperforms Fable 5, costs half as much, and continues to show that the sky’s the limit. GPT-5.6 xHigh running on Cerebras at 750 tokens/s is coming next week (promised for July by Sama), followed very soon by GPT-6, and presumably an updated version of Fable 5 as well. What an amazing time!
English
22
26
435
21.6K
Nick Masi
Nick Masi@Launch1Labs·
@rezoundous It's day by day, and we can't really capacity plan around any of it... Use it up as quickly and effectively as you can. Limits may reset, or the model may get pulled the next day.
English
0
0
1
97
Tyler
Tyler@rezoundous·
dare I say, Claude now has better limits than Codex
English
137
37
921
61.9K
Nick Masi
Nick Masi@Launch1Labs·
@taewan_02 Whats better... one super employee or a half dozen narrowly focused less powerful process executors?
English
1
0
1
56
C. Taewan
C. Taewan@taewan_02·
My new employeeee
C. Taewan tweet media
English
9
2
16
1.7K
Nick Masi
Nick Masi@Launch1Labs·
@signulll What is their pathway into number 3? What are they currently doing to bridge into the physical world? (Acquisition?)
English
0
0
0
4
signüll
signüll@signulll·
openai’s path to a $1T+ valuation is brutally f’cking simple. it’s three steps & new markets are just forming: 1) obliterate search & redefine consumer ai: ai-native search + personal agents that actually work will create entirely new markets & people will likely pay for them directly. google will have to dance & markets are going to crave this. 2) weaponize enterprise productivity: every knowledge worker is a few years away from being an ai-augmented force multiplier or a relic. businesses won’t just adopt a product like chatgpt, they’ll basically starve without it. 3) bring intelligence into the physical world: consumer & industrial robotics turn ai from a digital tool into a labor revolution. when machines think, everything changes. general-purpose robots will probably be the biggest wealth-creation event since the dawn of time. three inevitabilities, at least one trillion-dollar outcome—likely infinite if executed well.
English
99
62
1.1K
174.5K
Iman Mostafavi
Iman Mostafavi@imost·
@Launch1Labs Congrats! May not matter much since the GX10 has pretty good cooling, but they will run slightly cooler (1.4 C at load) if you flip it onto the other side (power button down). Might not be as aesthetic though.
English
1
0
2
30
Nick Masi
Nick Masi@Launch1Labs·
Second GB10 secured. This is what a future development team looks like: Two NVIDIA GB10-powered systems running large AI models locally. Mac Studio + Mac minis orchestrating agents. Raspberry Pis serving as always-on operators for monitoring, automation, and network tasks. UniFi using VLAN segmentation to isolate agent workloads and secure access across the system. The org chart is starting to look a lot more like a network diagram.
Nick Masi tweet media
English
4
0
8
312
Nick Masi
Nick Masi@Launch1Labs·
@alexisohanian Agency and grit are the most important characteristics to instill.
English
0
0
1
8
Nick Masi
Nick Masi@Launch1Labs·
@jefielding Don’t know what you meant by optimize for geo… And is ghost page different then SEO optimization?
English
0
0
0
3
Nick Masi
Nick Masi@Launch1Labs·
Early testing running Qwen3.6-35B-A3B in NVFP4 on 2GB10s. +28% to 97tok/s How am I doing?
Nick Masi tweet media
English
0
0
0
19
Nick Masi
Nick Masi@Launch1Labs·
@imPenny2x It’s artificial general intelligence. Yes, we already have AGI.
English
0
0
0
26
Penny2x
Penny2x@imPenny2x·
AI already drives safer than humans. It does math better than humans. It plays chess better than humans. It writes code better than humans. When Elon says AI will be smarter than the sum of all human intelligence in 5 years, he is being conservative. Just incase you haven’t actually processed it yet. AGI is here. It will compound and soon we won’t even be able to measure how much smarter it is than us without its help.
English
56
22
314
9K
Aravind Srinivas
Aravind Srinivas@AravSrinivas·
today is the most positive i have felt about the future of american ai. the coming together of so many companies in a united manner to fight against regulatory capture is incredible to watch. 🇺🇸
English
41
106
1.7K
116.7K
Nick Masi
Nick Masi@Launch1Labs·
Optimus will become the primary product. Most people don’t fully understand this yet.
Whole Mars Catalog@wholemars

If you're an investor who has lost faith in $TSLA, I can't blame you. This is a company that used to see record sales every year, 50%, 80%, 100% year over year growth. That's a distant memory for investors now, with the last year of record deliveries being 2023. You gritted your teeth and bit your tongue as you listened to your fellow investors cope. "Well, vehicle deliveries don't actually matter. Autonomy does" they said as the sales decline accelerated. "Amazing abundance!" they tweeted as Tesla posted its worst year of sales growth in the history of the company. So I can't blame you, I really can't, for having doubts about whether this company is headed for explosive growth. Some people who still want to believe may be rude to you as you exit, or call you a moron. I called people who can't see the potential morons yesterday. But I'm sorry about that, and would like to apologize and wish my former Tesla bulls off with love. You're not insane for feeling something is off. I'm probably the insane one for still buying. Maybe I'm just retarded, but I want to own more of this company. It could be blind hope, but based on everything I'm seeing it really looks like they're turning a corner and returning to growth. I mentioned that we haven't seen year over year sales growth since 2023. Last year, in 2025, Tesla delivered 153,097 units less than in 2024. For 2026, they've already delivered 117,346 more cars than they did in the first half of 2025. It looks likely that Tesla will post sales growth this year for the first time since 2023, and they've got a shot at 2026 being a new record sales year if they can deliver an average of ~485,000 cars in the second half of 2026 — just a bit more than what they delivered in the second quarter. My car is driving me around like a work of art. It's smooth and precise, sophisticated and refined, and the safety is off the charts. It can react faster than I can and spot things that I missed. They have the best self-driving system in the world with the best unit economics. The world has not woken up to this at all, but they will increasingly with each new Robotaxi city and each new FSD drive. Self-driving is finally starting to take off. For the first time, the majority (55%) of Tesla orders in North America ordered with the $99 a month self-driving option selected. 45% of paid self-driving users are now on the monthly subscription (as opposed to the legacy option of paying up front) and as more people move to the monthly subscription Tesla is about to cross $1 billion ARR from self-driving subscriptions. Self-driving is already a billion dollar business for them, and it's going to be a trillion dollar business. Then there's the driverless Robotaxi service. I just tried it in Austin. It's really hard to not get excited about the future Tesla is building when you can call a driverless Tesla and have it drive you around Austin for an hour while you watch Netflix or play games on the front screen. Then there's the Cybercab... the way they've hyper optimized this thing to be the lowest cost transportation option possible is absolutely refreshing in a world where the average marketed price of a new car is well north of $50,000. I think this is going to hit the market like a glass of ice water in hell. A $25,000 - $30,000 autonomous electric vehicles with no controls, that doesn't require any insurance — just a self-driving subscription. Optimus and digital Optimus could be much bigger than the entire transportation segment for the company. We've seen frontier models get proficient in controlling our computers. Why couldn't they control a humanoid robot too? I think it's very obvious to everyone that they will inevitably be able to. As that happens, a general purpose robot will be in every home and every business. Sometimes several robots. Tesla is designing everything from the chips to the manufacturing to the models to work together like an Apple product. If they can take even a small % of this market and I think they will do extremely well, because it's a massive market. But really, when you invest you're investing in the team. A lackluster team can take a great business and turn it into a struggling one, and a great team can take a struggling business and turn it into an incredible one (see: Apple). So the team is really the single most important thing to look at when buying any company. I believe in Elon and his team. They have been executing like crazy lately. I can't be the only one who has seen it. Elon didn't become the richest guy in the world by losing his investors money. When SolarCity had liquidity challenges and nobody was left to help, he made a deal with Tesla. When he was forced to pay $44 billion for Twitter despite his claim that the actual number of human users had been wildly inflated inflated, people thought investors were toast. Those shares are now publicly traded as part of SpaceX. I think Elon has a new plan to build a physical and digital AI giant to rival OpenAI and Anthropic in the race for superintelligence. So I don't know, maybe I'm just an idiot. But I love these products, and they make my life so much better everyday. I just can't shake the belief that the world is going to want them too. Everyone needs autonomous transport, and everyone needs a robot to help them sometimes. so the more $TSLA dips the more I'll buy. Maybe I'll lose all my money, but that's part of the fun. I just want the future this company is building. to all the other idiots who are still holding like me, we'll get through this together

English
0
0
1
34
Sam Altman
Sam Altman@sama·
i want the US to win in AI both in open source and proprietary models, and i am glad to see this
Jensen Huang@JensenHuang

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-Weigh…

English
1.8K
734
11.5K
2.4M
Jensen Huang
Jensen Huang@JensenHuang·
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-Weigh…
Jensen Huang tweet mediaJensen Huang tweet mediaJensen Huang tweet media
English
13.4K
22.8K
133.5K
40.7M
Nick Masi retweetledi
Nick Masi
Nick Masi@Launch1Labs·
@bindureddy do we know how big it will be? can it run on 2 GB10s?
English
0
0
0
825
Bindu Reddy
Bindu Reddy@bindureddy·
Kimi K3 will be truly open-weight and 3x faster on Monday Get ready to move all your standard workloads immediately Anything that is on Sonnet or GPT 5.5 can move — it's faster, cheaper, and better 🚀🚀
English
43
47
598
32.5K
Martin Tobias (Pre-Seed VC)
Martin Tobias (Pre-Seed VC)@MartinGTobias·
most pitch decks are 20 slides too long. if you can't explain your business in 5 slides, you don't understand your business.
English
30
5
118
16.2K