Moores Lawyer

A bored lawyer and AI enthusiast

[Moores Lawyer @newtemeraire]

@gabepereyra @harvey Interesting project. Will this be limited to post training and harness customisation?

[Moores Lawyer @newtemeraire]

We've seen this before in the early internet. Airlines, hotels, insurance etc. resisted aggregators and price comparison websites. However they eventually relented as that is where the customers were. I imagine the same will happen here - your agent will be able to use these websites (likely using agreed APIs over time) but companies will fight hard to have a direct relationship with them or at least for you to know you are using their product.

[Moores Lawyer @newtemeraire]

I agree that agents like muse pose lots of opportunities - but will most of value not fall to the agents and the business’s who actually produce things at source. An agent economy will cut out lots of the middleman who used to build apps. That is not to say there will not be value in creating dara bases that these agents use.

[Moores Lawyer @newtemeraire]

@quxiaoyin But Harvey does not capture any of its customer data

[Moores Lawyer @newtemeraire]

@CharlieGoes2War @harvey @andrewarruda Are you referring to when they post train open weight models? Do you think they data set will have been much better to what astra legal was post trained on?

[Moores Lawyer @newtemeraire]

@alexandr_wang When will it be available in the UK?

[Moores Lawyer @newtemeraire]

As others have pointed out there also sorts of strange incentives for lawyers. What is sometimes overlooked is the internal dynamics within firms. So much of being a good lawyer is not fucking up. The partners do not really care about your AI use but they will query why you did not take the approach that is tried and tested. Additionally getting AI set up takes time and for many they only thing they can see the deal they are on - if it does not save them time for that then they do not bother.

[Moores Lawyer @newtemeraire]

@CharlieGoes2War @harvey @andrewarruda But Harvey does not own any of the data that is the law firms. What is proprietary is how they utilise the models and their post trained open source models.

[Moores Lawyer @newtemeraire]

I think the real challenge for them is whether they can succeed without a machine of consultants (‘forward deployed engineers’) that law firms really value as lawyers are surprisingly inept about using any technology and are extremely conservative, thus they want their hand held on what to do.

[Moores Lawyer @newtemeraire]

@alexandr_wang When will it be available in Britain?

[Moores Lawyer @newtemeraire]

This is an interesting model from them. It avoids OpenAI having to get too involved in hand holding the customer like Harvey or Legora; but still gives some handholding and this startup allows the law firms to become slightly less dependent on their legal tech providers as they are building up domestic talent.

[Moores Lawyer @newtemeraire]

@totalwar The minor factions will just fill in the independent/rebel provinces which is a nice touch. However I hold my judgement till I see how much of a money grab this is.

[Moores Lawyer @newtemeraire]

@unusual_whales I wonder if this will break the trust that law firms might have to use these models as I presume these companies similarly had agreements not allowing their data to be used or observed.

[Moores Lawyer @newtemeraire]

@emollick This is not even factoring the costs of actually securing this data appropriately.

[Moores Lawyer @newtemeraire]

Jevons may mean vastly more law gets done without vastly more money going to lawyers.

The marginal volume AI unlocks is likely to be things previously decided were not worth paying lawyers to do: small claims, minor transactions, another layer of diligence to close off marginal risk.

AI can make those economical. It doesn’t create 10x as many high-value transactions.

And if AI is doing 90% of the new work, it’s not obvious why law firms capture most of the value.

[Moores Lawyer @newtemeraire]

We will see more mega law firms doing this in the coming year or so. It is good marketing. However their tuning of the open source models will not be better than any others as they are fundamentally not a tech company and the gains from their data will be marginal + it seems unlikely their servers are that much more trustworthy than some of the large cloud providers.

[Moores Lawyer @newtemeraire]

It should identify the text I have asked to be edited in quotation marks or after a colon, and then use whatever you already apply to incremental iterations of outputs that Harvey produces

[Moores Lawyer @newtemeraire]

@ChatGPT Is this just their response to Grok Bot?

[Moores Lawyer @newtemeraire]

Grokbots connector to google docs does not allow it to edit files but can create new ones - why is that?