Sabitlenmiş Tweet
Joe Botsch
5.4K posts
Joe Botsch
@jbotsch
VC @Deviationcap, formerly @rtpvc, @OpenView & Head of Research at @MassChallenge. talking startups or complaining about Boston sports
New York, NY Katılım Temmuz 2011
1.9K Takip Edilen861 Takipçiler
Such a fun interview!
Dwarkesh Patel@dwarkesh_sp
New blackboard lecture w @ericjang11 He walks through how to build AlphaGo from scratch, but with modern AI tools. Sometimes you understand the future better by stepping backward. AlphaGo is still the cleanest worked example of the primitives of intelligence: search, learning from experience, and self-play. You have to go back to 2017 to get insight into how the more general AIs of the future might learn. Once he explained how AlphaGo works, it gave us the context to have a discussion about how RL works in LLMs and how it could work better – naive policy gradient RL has to figure out which of the 100k+ tokens in your trajectory actually got you the right answer, while AlphaGo’s MCTS suggests a strictly better action every single move, giving you a training target that sidesteps the credit assignment problem. The way humans learn is surely closer to the second. Eric also kickstarted an Autoresearch loop on his project. And it was very interesting to discuss which parts of AI research LLMs can already automate pretty well (implementing and running experiments, optimizing hyperparameters) and which they still struggle with (choosing the right question to investigate next, escaping research dead ends). Informative to all the recent discussion about when we should expect an intelligence explosion, and what it would look like from the inside. Timestamps: 0:00:00 – Basics of Go 0:08:06 – Monte Carlo Tree Search 0:31:53 – What the neural network does 1:00:22 – Self-play 1:25:27 – Alternative RL approaches 1:45:36 – Why doesn’t MCTS work for LLMs 2:00:58 – Off-policy training 2:11:51 – RL is even more information inefficient than you thought 2:22:05 – Automated AI researchers
English
@nedoleary That looks like it’s gonna get in the way of the distillery and lord of the rings theme park I want to open…
English
if you're 25-30 and your main circle isn't frequently evaluating vast tracts of picturesque undeveloped land for new data centers, it's time to elevate your circle
English
Joe Botsch retweetledi
BU's all in tuition costs for the 2026-27 academic year is projected to be between $95K and $98K, so I guess we are celebrating Green virtue signaling so we can start kids out in life with $400K in debt?
Only In Boston@OnlyInBOS
Boston University's new 19 story building runs with: No gas No fossil fuels No emissions One of the greenest buildings in New England.
English
@nikunj The competitor investing is going to get worse especially with the neolabs, lots of former teammates starting competing companies
English
Here’s the unsaid truth about venture investing..
Most* large AUM fund can’t afford to miss a category to generate the returns they need to justify now. This means they WILL invest in competitors. They might invest in different stages or sometimes at the same stage. And, because they generally invest in many seed companies, portfolio companies will end up pivoting into the same space.
This rightfully pisses founders off since they get one company and funds get a portfolio. But that is the risk you should be wary of when you’re choosing your capital partner.
Loyalty is already on very thin ice in tech and this is going to get greatly exacerbated in the next few years. Expect more noise and volatility.
At the end of the day, capital and the venture partner are a tool in the founders arsenal. And, they single-handedly won’t make or break your company. Regardless, choose them wisely!
* Before the pitchforks arrive, I’m NOT tooting my own horn or saying large AUM funds are bad. Some (and slowly shrinking pool) large AUM funds are still extremely concentrated and loyal. Just stating the facts as I see them on the floor today.
English
Joe Botsch retweetledi
First blog post up on Robotics Simulation Infrastructure! I give a high-level overview, followed by an elementary example of better infrastructure for pose management. Link in thread below
Stone Tao@Stone_Tao
would people be interested in a blog series on "simulation infrastructure" where I look at popular robotics sim tools (e.g. mjlab, isaac...) and explain the pros and cons of their design decisions. The series would cover what I believe to be key pillars of good, usable, sim code
English
Joe Botsch retweetledi
Big announcement today at @remote!
We are going all in on 𝗴𝗹𝗼𝗯𝗮𝗹 𝗽𝗮𝘆𝗿𝗼𝗹𝗹 𝗮𝗻𝗱 𝗲𝗺𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲, and opening our platform to the world.
Sounds simple until you try to build it:
Payroll in one country is already a dense topic. Payroll across the world becomes an insane systems problem: tax laws, social security, benefits, holidays, languages, currencies, equity, and local interpretations that change because a government updated a rule somewhere.
When we started Remote, my engineer brain thought we would plug into the payroll APIs that existed around the world.
They mostly did not exist, so we did the hard work.
We set up entities. We built in-house payroll engines. We documented compliance country by country. We learned the local details that make payroll work in practice.
It took seven years. It took a lot of investment. It took a lot of people doing important work.
Today we are opening that infrastructure up.
Any company, any tool, any workflow, any AI agent can plug into Remote and use the global payroll and employment infrastructure we have built.
You keep your systems. We handle the hard parts: payroll, compliance, employment infrastructure, country by country.
I am grateful to our team, our customers, our partners, and everyone who supports us along the way.
Seven years building toward this chapter. The best is yet to come.
Let’s go.
Full story: okt.to/VOw785
GIF
English
Joe Botsch retweetledi
The skilled trades shortage didn't happen to America. America chose it.
Dynamo's @santoshsankar on the 1983 report that dismantled the vocational pipeline, and the $325.6B annual bill arriving now.
Read the full piece below:
buff.ly/74nA5UL
English
I raised my personal fund without a formal deck, thesis doc, or pitch.
Just texted a handful of our existing investors over the weekend, and money was wired into the account within 2 hrs.
Some of them were so excited about the idea that they asked if they could bring another friend.
English
Remember the kid who reminded the teacher that they didn’t assign homework?
shirish@shiri_shh
This tweet single-handedly made Anthropic and OpenAI scramble and issue full statements on secondary stock deals. $500B of paper value just got wiped out overnight.
English
Arguably *the* must attend annual event for GPs and LPs alike
@Rick_Zullo and team created and continue to cultivate this thriving community of GPs building firms in thoughtful and different ways to support founders
Read more below and apply to participate
Rick Zullo@Rick_Zullo
1) I’m excited to announce @EqualVentures’s 5th Annual Emerging Manager Circle Summit 8 years ago, this was just a few of us swapping stories, trying to help each other. I couldn’t imagine what EMC would become Now, we need events like EMC more than ever newsletter.equal.vc/p/announcing-t…
English
Joe Botsch retweetledi
If anyone builds it, everyone thrives. Over the past decade, a lot of important work on AI alignment has focused on avoiding harm. But freedom from harm isn't the same as freedom to flourish.
In this paper, we introduce 'Positive Alignment'. A positively aligned agent is one that helps us navigate our own value trade-offs, builds our resilience, and acts as a scaffold for human flourishing. Doing this without slipping into top-down, technocratic paternalism is the great design challenge of our time.
We think a lot more research is now needed to explore this frontier: how do we align models that actively help us thrive?
Amazing work by @RubenLaukkonen, @drmichaellevin, @weballergy, @verena_rieser, @AdamCElwood, @996roma, @FranklinMatija, @shamilch, @_fernando_rosas, @scychan_brains, @matybohacek, @sudoraohacker, and others.
arxiv.org/abs/2605.10310
English
Joe Botsch retweetledi
NFL SCHEDULE LEAK
Jets at Dolphins - Week 1 3:00 AM on Cartoon Network (Sun Sep 13th)
English
Joe Botsch retweetledi
The Inference Shift
Agentic inference is going to be different than the inference we use today, and it will change compute infrastructure because speed won't matter when humans aren't involved.
stratechery.com/2026/the-infer…
English
pre-kids: red eyes are the worst - why would anyone take them??
post-kids: red eyes are honestly the best. i can tuck my kids in bed and wake up in a new city and get going right away.
(typing this from a red eye headed to NYC rn)
English
Joe Botsch retweetledi
Neural networks might speak English, but they think in shapes.
Understanding their rich *neural geometry* is key to understanding how they work – and to debugging and controlling them with precision.
Starting today, we’re releasing a series of posts on this research agenda. 🧵
English
@bznotes It’s not what you don’t know, it’s what you know for certain that just ain’t so!
English
You can never go wrong investing in consensus bets :-)
Rory O'Driscoll@rodriscoll
Whenever something looks incredibly easy, and it becomes the conventional wisdom that everyone's going to make money, that’s usually when it blows up in your face. In 2019, The WSJ was writing that the best PE firms “never lost money.” The deals that those firms did right after are now blowing up. Overconfidence is a fatal error in investing. Something we probably should remember. wsj.com/articles/vista…
English