Robert Wachen
@robertwachen
Followers
2K
Following
551
Media
5
Statuses
225
cofounder @etched. prev @harvard; co-founded prod, mentor labs (acq.)
Joined March 2014
Paradigm has saved me from many late nights of spreadsheet work - strongly recommend.
Paradigm is the AI-native spreadsheet to eliminate menial work. Thousands of users have saved 10,000+ hours with Paradigm, and you can be next. Get your first month free today, then plans start at just $20/month.
0
1
39
See you there! I’ll be judging.
Last call to apply for the @MARA Hackathon. Solve AI + energy challenges. Compete for 1 BTC. June 21. San Francisco.
0
3
5
Blazing-fast image creation – using just your voice. Try Grok Imagine.
284
564
3K
July 14, 2023 (~19mo ago), I got a call from @suryamidha (pictured),.@BrendanFoody, and @adarsh_exe to come to their Menlo Park office ASAP. They had just closed their first major contract and were debating whether to drop out. Adarsh, who had turned down his Bridgewater offer
Mercor is solving talent allocation in the AI economy. The difference between greatness and failure is the right person being in the right place at the right time. Putting them there is the hardest unsolved problem in capitalism. We’re excited to announce our $100M Series B at
3
5
125
Come hang! $60k in cash prizes and each team gets their own 8xh100, gonna be nuts.
We're excited to partner with @Cognition_Labs @Mercor_AI @CoreWeave and @AnthropicAI to host an inference-time compute hackathon, featuring >$60K in cash prizes and >1 exaflop of free compute.
0
2
28
As long as compute and intelligence are correlated, specialized chips are inevitable.
New verified ARC-AGI-Pub SoTA!. @OpenAI o3 has scored a breakthrough 75.7% on the ARC-AGI Semi-Private Evaluation. And a high-compute o3 configuration (not eligible for ARC-AGI-Pub) scored 87.5% on the Semi-Private Eval. 1/4
0
1
20
The future is here @walden_yan @SilasAlberti.
Devin is generally available today! . Just tag Devin to fix frontend bugs, create first-draft PRs for backlog tasks, make refactors, and more. Start building with Devin below:
1
1
16
GPUs started for graphics and now are used for transformers . Maybe Sohu will start for transformers and end up doing graphics.
Introducing Oasis: the first playable AI-generated game. We partnered with @DecartAI to build a real-time, interactive world model that runs >10x faster on Sohu. We're open-sourcing the model architecture, weights, and research. Here's how it works (and a demo you can play!):
6
0
31
Been crazy to watch Mercor’s journey. Huge congrats and excited for what’s ahead @BrendanFoody @adarsh_exe @suryamidha.
Mercor is solving global labor matching with models that understand human ability. @mercor_ai raised a $30M Series A at a $250M valuation, led by @victoralazarte and @bgurley at @benchmark, with participation from @peterthiel, @jack, @adamdangelo, and @LHSummers.
4
4
28
Was a lot of fun - tune in for a deep dive on all things AI chips.
INTERVIEW WITH FOUNDERS OF @ETCHED.the thiel fellows fighting nvidia. - @UbertiGavin & @robertwachen on AI chips.- Debating future unemployment predictions.- AI's impact on long term GDP growth.- Raising their $120m Series A.- Slow vs Fast Takeoffs.- much more.
1
0
55
We’re living in the largest infrastructure buildout of all time. The chips won’t make themselves - excited to bring Sohu to reality.
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,
11
7
202
"Even if there were no algorithmic improvements from here, . the scaling laws are going to continue"
0
1
5
Last year Dario predicted billion-dollar training runs this year and up to $10B training runs in 2025:
1
1
6
Someone should make an AI agent to automatically send 3-way NDAs to every existing vendor every time you add a new vendor.
0
0
7
Hundreds of millions are poured into projects like Humane expecting real-time AI to be around the corner and are stuck waiting for the hardware to catch up. We’re going to need a breakthrough in computing to power massive, high-batch, real-time AI models.
0
1
4
Most compute cost decreases are from:.- Companies deciding to price their tokens at cost.- VCs subsidizing loss-leader inference-as-a-service pricing.- Cloud companies selling cheaper GPU hours due to overoptimistic depreciation cycles (ex. 7-8 years).
1
0
7
Hot take: silicon valley assumes that token costs and latency will magically keep dropping by 10x. Unfortunately GPUs aren’t getting better fast enough.
It is an incredibly safe assumption that models get much cheaper and smarter from here, yet very few founders are acting like it.
3
0
31