Head of AI @ Keller Postman | former Head of Legal @ Crosby
New York, NY
Joined March 2009
- Tweets2.7K
- Following771
- Followers2.5K
- Likes10.7K
Things are about to get very strange
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
I muted the phrase “holy sh*t” and nuked like 80% of the AI clickbait slop from my timeline. Highly recommend
Today @dots caught that I booked a the wrong hotel dates for a conference — my check out was a day before the conference ended (and a day before my return flight!)
I think this would not have been caught even 4 weeks ago
Instinct is way ahead on group experiences in the personal agent game
@Muse hasn’t shipped parity to even the “trusted people” feature, and now they added an agent directly in a group chat
Shared agents carry with the most risk, so I understand Meta’s caution, but imo this is a huge differentiator and enough to keep me from churning off Instinct
Instinct in group chats
Starting today, early access users can add Instinct to group chats. A new Instinct joins and works for the whole group.
Making plans with friends usually turns into a frustrating back-and-forth over times and places. With Instinct in the group, you can explore options together, agree on a plan and get it done, all in one thread. Your friends don't even need Instinct to join in.
A few things it’s good at:
- Planning a weekend trip with friends, including dates and arrival times
- Coordinating logistics with a roommate
- Getting tickets as soon as they go on sale, then splitting the cost and sending one to everyone
- Running a fantasy league, keeping standings, settling rules disputes, and reminding everyone to set their lineups
- Sorting out carpool schedules - works out who’s driving each day from everyone’s schedules
- Organizing Thanksgiving, keeping track of who’s bringing what and ordering the shopping list
Group chats are new territory for Instinct, and we put a lot of care into getting the permissions right. Here’s how it works:
- Your personal Instinct asks before connecting with the group’s Instinct. You choose which groups to trust and can remove trust at any time.
- The group’s Instinct has no direct access to your personal accounts. Your Instinct checks your permissions before sharing information or taking action.
- If someone new joins the group, all pending replies from your Instinct are held until you grant it permission to share with the expanded group.
Group chats are available to all of our early access users, and will be rolling out to everyone soon. To join early access for this release, ask your Instinct to put you on the list.
Great work with @GajanNagaraj @anuda_w and @saagarwashere
I am not impressed by Grok bot
Grok is running a noticeably worse model than Astra. The multi bot thing is not really helpful — it’s a strange abstraction that gives you a dopamine hit for making a new bot but doesn’t produce noticeably better results
The only advantage is that it’s easy to send X bookmarks to Grok bot but I don’t think that is useful enough to maintain a whole separate sub
What am I missing? I thought this was supposed to change my life
Just heard “that’s not leadership, that’s a puppet” in a political TV ad
AI slop has reached the super PACs
A lot of the news on misalignment seems to be coming from OAI
It sort of feels like Anthropic wants to pace the frontier because they think OAI has been reckless playing catchup
It’s not really “pace the frontier” but “someone needs to stop those guys”
Exclusive: OpenAI is scrapping the release of its next-generation AI model because it failed to meet safety standards on.wsj.com/4xPPgao
I love that S&C partnered with OAI on something as generic as “agreement analyzer”
If you ever feel like you’re behind, headlines like this can give you hope
More law firm/foundation model lab collaborations; this time in a post at Legaltech News (LTN) about Sullivan & Cromwell's collaboration with OpenAI to build the firm's proprietary "Agreement Analyzer."
From the LTN post:
"A growing number of law firms have launched software co-developed with Big Tech and legal tech companies in recent months as part of broader efforts to both deepen vendor relationships and differentiate through the use of artificial intelligence.
"One such firm is Sullivan & Cromwell, which on Sept. 17 announced the development of a bespoke Agreement Analyzer built with OpenAI. The Analyzer's output includes memoranda which list issues in the proposed agreements, provide the rationale for the issues, and offer a set of proposed treatments that would address the problems. It also provides a Microsoft Word document with tracked changes to the agreement, which users can accept, modify or reject."
$$Quote:
"[Sullivan & Cromwell partner, John] Savva designed an initial proof of concept for the too, which was turned into a working prototype with assistance from the firm's IT staff in roughly a week. Sullivan & Cromwell then turned to OpenAI to refine the tool and prepare it for wider use.
"'We were able to take this proof-of-concept tool that we had initially developed, provide the code to OpenAI, [and] work with our engineers to do a few things,' he said. 'One was improve overall performance. Two, in particular, look for ways to speed it up. ... We worked with the OpenAI engineering team to find opportunities to have processes work in parallel, work faster, work more efficiently.'"
Do read the entire LTN post at the following link to get an illuminating account of how Sullivan & Cromwell worked with OpenAI to build the AI tool discussed here and also to get an idea of some of the broader implications (for law firms, their lawyers and lawyering itself) of these law firm/foundation model collaborations: law.com/legaltechnews/2026/0…
#legaltechnology #legaltech @sullcrom
Harvey was a trying to follow Cursor’s path. I think it’s more defensible than remaining a pure wrapper company
Everyone is sharing this to dunk on Harvey but I think that misses the mark entirely
Cursor was burning cash by subsidizing token costs on their platform. Their absolute to post train a coding model was a huge part of why they got acquired for $60B
There are lots of salient differences between coding and law. But cursor showed that the combo of good product sense and world class post training can create a very attractive outcome
Of course, it’s not clear that the legal vertical is as attractive to the labs for an acquisition as cursor’s data was. Still, subsidizing tokens and then post training provided the most favorable outcome of this AI cycle.
This is a good bet by Harvey
Trust Network is the only reason to use Instinct over Muse
I expect Meta to match this within a few weeks. But even if it doesn’t and users have time to create a big network within Instinct, it’s still just so easy to move from one agent platform to another. The agents themselves make it super easy if you just ask for a download of what they know about you
When compute is so expensive, and compute is the durable advantage, gonna be hard to beat Meta in any consumer app
Is “Harvey for supplement formulation” bullish or bearish for Harvey
Brilliant business no one is talking about:
IngredientAI: AI for supplement formulation.
Think Harvey but instead of being for legal it’s for supplement formulation.
I tested product and immediately invested. Here’s why:
1. 99% of supplements are filled garbage ingredients and no efficacy
2. Supplement manufacturing is hard
3. One of the FEW defensible use cases of ai against Claude: Claude likely will never have the data info the have and the expertise to be able to formulate them order supplements
If your in ecom pay attention
There’s never been a better time for high end knowledge work and high performing professional services firms.
Open, self hostable frameworks like Buzz and Hermes make it easy to customize your harness, build multiplayer capability, and capture proprietary data. You can even find best in class evals by just lurking on X.
The firms with competitive differentiation have an opportunity to compound their unique advantages, growing top and bottom line.
Firms that compete on price can win, too, but it’ll be bloodier since they’ll have to cut costs more aggressively (ie layoffs).
What are we doing at this point
Breaking news: Scientists in the US have for the first time used artificial intelligence to create viruses unknown in nature, a milestone that promises advances in healthcare but also raises important biosafety and biosecurity concerns. ft.trib.al/DmuWd4p
