Seattle
Joined March 2013
- Tweets316
- Following502
- Followers515
- Likes492
Pierre Martin retweeted
I've come up with a novel idea for how humanity should deal with AI risks.
It's powerful and deceptively simple. In fact, it can be stated in one sentence.
And it relates to something I learned from Elle Woods. (thread 🧵)
Pierre Martin retweeted
Hey Laura, thanks again for drawing attention to Sway, a tool anyone can use to promote or support any viewpoint.
If you're trying to attack me, can you try harder? Your jabs aren’t even landing. Let me help you.
I’ve supported both Republicans and Democrats in the past, but way more Democrats. I donated to the campaigns of Obama, Hillary, Tulsi Gabbard, and others. Then I mostly stopped donating to politicians, but I tweeted many times in support of Biden, which was a mistake. Biden’s presidency was a huge disappointment.
I’ve made many other mistakes. I’m often wrong, and I’m not afraid of admitting when I’m wrong. I’m not afraid of anything. I’m certainly not afraid of you.
That said, Sway isn’t about me or my viewpoints. I don’t even have my own page on it. Sway is technology, and good technology helps everybody. Sway never supports one viewpoint over another.
I was glad that MTG promoted Sway, even though I don’t agree with all her ideas. She’s entitled to her own viewpoints, and so are you. One thing I love about America is that all viewpoints are allowed, and the same is true on X, and on Sway.
When you tweeted about me earlier today, I replied that we’ll work with people all across the political spectrum, including you. I meant it.
Sway simply offers tools for people to advance the causes they care about. In fact, even in the video you just shared, I tell the anti-ICE guy: “someone could also create a group that is to strengthen ICE on Sway and try to make their case to voters as well.”
Sway has voter guides on many subjects. Anybody can make one. For example, here’s a Trump supporter using Sway to promote a conservative voter guide for Californians: instagram.com/reel/DXiQBO6CY…
I hope that helps. Thanks for all the entertaining tweets today Laura, and I just followed you so that I don’t miss the next one!
Everybody dismisses the AI "wrappers" until they try to build one.
👇 Practical experience from building @Gavel_io, turning data into real legal work product, and how I foolishly dismissed @cursor_ai as a wrapper in 2023.
open.substack.com/pub/pierre…
Pierre Martin retweeted
Our COO @mauricioduartel sat down with @Gavel_io CTO @PierreMartin_7 to talk AI in legal tech. This conversation helped inspire Pierre's latest piece, which explores how AI and legal training complement each other.
Read the full article: pierremartinai.substack.com/…
#legaltech
ALT Full article on Substack
I expected lawyers to be behind on AI & I was wrong.
I came to know a lawyer very well over the course of building @Gavel_io with @dorna_at_gavel . For the past 3 years people have heard me blabbering about how engineers and lawyers are wired the same. In rough lines, both types are geeks with a passion for details and poor sales skills. In practice, the similarities in how we work are striking. There is no surprise computer code and legal code are the applications that are seeing the fastest adoption of AI. My newest substack explores this fact in more details: pierremartinai.substack.com/…
I expected lawyers to be anti AI. They are not. They are anti hand-waving. And so are engineers.
So the path to adoption means that legal AI must be grounded in their documents, their precedents, their workflows, and their professional judgment. We are building these harnesses directly inside Gavel Exec and @Microsoft Word - helping lawyers move faster with AI, and without asking them to lower their standards.
That distinction matters. Would love your thoughts.
Have you heard of supply chains? They have been seriously disrupted for the past 6 years, and I know something about it having been involved in building physical supply chain technology at Amazon and leading startups during both hypergrowth phases and the height of the pandemic or the Suez canal shut down. The key to effective supply chain operations is always deep visibility into all actors involved in the supply chain. And technology is the key to achieving this visibility and ensuring that everything stays under control.
But did you know there is such a thing as Software supply chains, and AI supply chains? For instance, when you sign a contract with any legal AI vendor, you are also implicitly engaging many vendors in their digital supply chains: whether OpenAI, Anthropic, Amazon Web Services (AWS), Google Cloud, Microsoft, Cloudflare, database providers, infrastructure providers etc. etc. The bytes, and your data, are moving through many partners, and almost no one tells you that in a demo.
At Gavel, we use many of these solutions in our AI supply chain, as you'd expect. So does Harvey. So does any serious legal AI company. Having an AI supply chain is not a red flag. But not disclosing it, or researching it, is a mistake.
So as lawyers evaluating solutions, ask for the map. Ask how your data can be used by anyone in the supply chain. Use the inquiry skills and critical thinking you've acquired in law school.
I wrote the full evaluation framework in my Substack. open.substack.com/pub/pierre…
#legaltech #legal #technology #ai
#Gavel #BuiltOnGavel #AccessToJustice
Pierre Martin retweeted
There are 9 days left to apply to Neo Residency (for startups & students).
We've already accepted more than a dozen teams, and there are a few slots left.
I can't wait to spend the summer with this incredible group.
It’s the final stretch! Neo Residency Summer 2026 applications are due April 30th at 11:59pm [PT] 🎯
Apply now at neo.com/residency
We gave a group of practicing deal attorneys a contract to redline. Then we gave the AI the exact same contract and compared results.
We measured precision and recall, typical indicators used to benchmark search and ranking algorithms. In our case recall means: how often did the AI catch the same issues the human caught? Precision means: how often did it flag something the human did not?
As we improve the AI, we are actually seeing precision decrease. When you look into it, it is not because of hallucination. It is because AI is picking up inconsistencies in large documents that a human actually misses. Turns out machines have an unfair advantage in reading 100s of pages and retaining all of it in short term memory.
Full piece: lnkd.in/gEUZUwbv
Most legal AI benchmarks (or any AI benchmarks really) measure model intelligence. Legal buyers should care about production reliability. Those are not the same thing.
My first newsletter is a reality check on what to actually ask when evaluating AI, after 17 years of building and deploying it in healthcare, logistics, and legal.
Read it and subscribe:
pierremartinai.substack.com/…
17 years deploying AI in the real world (healthcare, logistics, law), the pattern is always the same: turn human expertise into systems.
Lawyers who win will be the ones who turn judgment into reusable workflows, memory, and precedents
Knowledge without systems loses in the end
Spot on! lnkd.in/gMuaQeZA
In the push to shift feedback loops left, my most helpful harness of left is an agent to review a plan for common system issues & YAGNI. It's become really good at challenging even the looks-genius-but-actually-stupid ideas of yours truly.
Pierre Martin retweeted
Yann LeCun (@ylecun ) explains why LLMs are so limited in terms of real-world intelligence.
Says the biggest LLM is trained on about 30 trillion words, which is roughly 10 to the power 14 bytes of text.
That sounds huge, but a 4 year old who has been awake about 16,000 hours has also taken in about 10 to the power 14 bytes through the eyes alone. So a small child has already seen as much raw data as the largest LLM has read.
But the child’s data is visual, continuous, noisy, and tied to actions: gravity, objects falling, hands grabbing, people moving, cause and effect. From this, the child builds an internal “world model” and intuitive physics, and can learn new tasks like loading a dishwasher from a handful of demonstrations.
LLMs only see disconnected text and are trained just to predict the next token. So they get very good at symbol patterns, exams, and code, but they lack grounded physical understanding, real common sense, and efficient learning from a few messy real-world experiences.
---
From 'Pioneer Works' YT channel (link in comment)
Let's goo!!
We've raised $2.3B in Series D funding from Accel, Andreessen Horowitz, Coatue, Thrive, Nvidia, and Google.
We're also happy to share that Cursor has grown to over $1B in annualized revenue and now produces more code than any other agent in the world.
This funding will allow us to invest deeply in research and build Cursor's next magical moment.
The team is stacked and bringing manufacturing to the age of AI. Customers are saving millions instantly, traction is out of the charts.
If you are not considering joining the movement, you are already missing out.
Congrats AJ, Leonie! Glad to b backing you from the start !
My cofounder @AnnaJuliaStorch and I grew up in manufacturing towns while spending our days chasing speed - on the mountains and the racetrack. With @DryftHQ , we bring our love for both worlds to life.
That’s why we couldn’t be more excited to finally come out of stealth and announce our $5M round led by @GeneralCatalyst, with participation from @Neo, @SherylSandberg ’s Fund, @JeffAWilke, @CHughesJohnson, Dr. Markus Flik, @Qasar and many other incredible Angels. (see below)
Pierre Martin retweeted
Trying something new:
Join my nationwide voting group to release the Epstein files: Sway.co/massie
I co-led the bipartisan bill to release the files. Now let's make Congress act.
When you commit to vote together you can influence anyone who wants your votes.
Join me.
