Deepseek is better for hardware production related questions. ChatGPT always takes the perspective of a detached sourcing specialist.
English
Michael Mayer
1.6K posts
Michael Mayer
@micjm
Co-founder of @bottomless Building the auto restocking home.
Seattle, WA Katılım Mart 2019
94 Takip Edilen3K Takipçiler
Hardware design is just software. On the other hand, hardware production is a full contact sport
English
I find myself pushing the team to use LLMs often, but one area it fails is when trying to source diverse feedback when everyone works with ChatGPT on their feedback.
English
Enterprises are often dissapointing until you meet the org most responsible for company success.
Suggests a power law in terms of the importance of various competencies.
English
This is super cool. A customer used our Pet Beta Program to hack their own smart pet feeder w/ access control. (Our smart scale is under the automatic feeder).
English
@matthucius You already shared the way, I think. There is only the present. The future and past don't exist other than memory and projection. Trying to change the past or future are futile.
Focus exclusively on what's in your control, in the present.
English
I am genuinely really scared of getting older, been having intense dreams/nightmares about it and shit - seems very unreasonable but also very real -
Like I am worried I’ll run out of time and life is passing by too quickly without achieving enough or maximising all experiences etc, hate to see parents getting older, friendships and chapters ending, get overly nostalgic about the passage of time day to day and how everything is just a memory the second it’s happened
Is there a name for this and is it common? I’m truly analysing it so much internally, or maybe I’m just a chronically deep thinker, but all of life and speed of time seems to be getting exponentially faster as of late 🤔🕰️🔮
English
Last year I met @UbertiGavin and had a series of fascinating conversations on future of AI hardware and software (including one on the podcast that I still recommend). His company, Etched, had set out to build the first ASIC for transformers that does one thing and one thing only: run lightning fast AI inference to enable any AI app you can think of that current speeds make impossible, because things run too slow and/or too expensively.
Today, I’m excited to share that Posive Sum has co-lead, with Primary Ventures, their $120M Series A financing, alongside Peter Thiel, Stan Druckenmiller and many other great investors
I highly recommend reading this announcement thread and memo and then start imagining what AI apps will be able to do when they are orders of magnitude faster and cheaper than even the next gen NVIDIA GPUs—real time video, real time voice interactions, agents and everything else we are all so excited about as users.
I think Etched has a chance to be an incredible enabler of people’s ideas, products, and companies.
Etched@Etched
Meet Sohu, the fastest AI chip of all time. With over 500,000 tokens per second running Llama 70B, Sohu lets you build products that are impossible on GPUs. One 8xSohu server replaces 160 H100s. Sohu is the first specialized chip (ASIC) for transformer models. By specializing, we get way more performance: Sohu can’t run CNNs, LSTMs, SSMs, or any other AI models. Today, every major AI product (ChatGPT, Claude, Gemini, Sora) is powered by transformers. Within a few years, every large AI model will run on custom chips. Here’s why specialized chips are inevitable:
English
@patrick_oshag @RobertGreene I think we got a few glimpses behind his mask.
English
A very interesting conversation with @RobertGreene, the author of the 48 Laws of Power, Mastery, and many other books
I was able to revisit my notes from 10+ years ago when I first read the books and ask him about my favorite sections
Makes me want to interview more authors!
English
@gokulr publicly traded investments are reversible, so this makes sense for him but not for venture
English
Reply to a low battery alert:
(Don't worry Michele, we won't leave you...)
English
@austin_rief @EricJorgenson I can see four quadrants to tweet about here
English
@EricJorgenson Have you thought about tweeting about working on the weekends in your 20s?
English
If you don't have enough followers, be sure to start a clumsy debate about VC vs Bootstrapping.
Be as opinionated, non-nuanced, and unhelpful as possible to incite max comments and quote tweets.
This has worked since the dawn of Twitter.
English
@patrick_oshag FWIW, capturing data yourself from your product's usage is an interesting moat and example of @bgurley's idea of the product getting better with more users.
ChatGPT has a critical mass of user ratings on responses. @bottomless is literally generating proprietary training data
English
@patrick_oshag data is FINALLY the new oil
AI will be able to do roughly anything that is currently done at scale, with data capture (cheaper and faster) and roughly no more than this.
English
This is a note on our internal chat about a specific AI application company, but amazing how much it applies to almost every AI app company we see.
“Some thoughts
- the LLM / RAG thing seems undifferentiated and won’t last
- The data is everything, so mostly curious to know their model for getting uniquely good data. Without that, will be brutal. Can’t just liscence from [redacted] etc.
- can we ask for access to the product before the call to make sure it’s real?
- I’m pretty impressed by that traction”
English
Would you prefer a smart container over a smart scale for replenishing coffee / protein powder / dog food / etc?
English
@patrick_oshag requirements: deep market, participants with different time horizons.
It would probably have to be something new, or newly legible.
Probably GPUs.
Maybe residuals on high quality training data. Some text corpuses should be more valuable than most recorded music, for example.
English
@patrick_oshag @InvestLikeBest @bgurley @mjmauboussin @ncsh @dittycheria @artofinvest excited for @dittycheria, became small shareholder after using their product and realizing we would never rip it out
English
A sample of conversations to come on @InvestLikeBest
- Marc Lasry
- @bgurley and @mjmauboussin together
- @ncsh
- @dittycheria
- 2 hours of Mitch Rales (on @artofinvest)
English
@zackkanter @firstround The written content they put out around this was pretty good
English
This looks awesome. Seems like few people are better positioned than @firstround to do this.
Brett Berson@brettberson
There isn’t a paint-by-numbers kit for any part of company building. But when it comes to product-market fit, founders have very little to go on. It’s seen as more art than science, with squishy definitions, no real data, and mostly high-level advice. For the past few years @firstround, we’ve been working on a big project to change that — and it launches today. Introducing PMF Method, an intensive 14-week experience designed to help exceptional founders build epic B2B SaaS companies. Along with our nearly 20 years of data from 500+ pre-PMF investments, my partner @tjack has personally talked to hundreds of founders on this topic. What emerged was a very consistent set of patterns for sales-led B2B companies (consumer and bottoms-up SaaS is very different in our view, a bit more alchemy involved). But in enterprise, we believe it’s possible to reduce the role of luck. We’ve distilled those patterns into a brand-new, detailed framework, and PMF Method’s 8 tactical sessions, where we help early founders discover what customers really want, build the right v1 product, and close their first enterprise sales. Applications close May 7th, but if you’re curious to learn more about our approach, we’ve decided to publish the framework that we cover in the first session in a new long-form essay — as we do with programs like The Review and Angel Track, we default toward openly sharing as much as we can with the broader startup ecosystem. You’ll find tons more details in the essay (linked in the next post), but in a nutshell, we break PMF down into 4 distinct levels, sharing detailed benchmarks to aim for, case studies and actual data from how Looker progressed through these stages, the signs that you’re getting stuck, and tactical advice from from incredible founders like @christinacaci, @zachperret, @lloydtabb, @jboehmig, and @jaltma. Check out the links below for more details. Can’t wait to read applications!
English
I truly think the world would be better if listening to Ken was required
I probably laughed more in this episode than any other
Patrick OShaughnessy@patrick_oshag
Ken Langone’s life should be a movie He co-founded Home Depot Bought and held shares for decades in others like Eli Lily Fought Spitzer (hilarious section) Rebuilt NYU Langone So much more Integrity and energy off the charts. The American Dream: joincolossus.com/episodes/32603…
English
@kirbywinfield Yeah, that makes sense!
Even the mega fund filling out the CRM feels fine... the associates are usually really bright, ambitious people who are learning to "think like VCs". It's usually interesting to speak with them.
English
