Rosie Teo

@mil000 Pretty much every user will need to think about token usage - even enterprise users - just for that, model selection is useful

If you app has a AI model picker that lets the user select between Claude, ChatGPT and maybe Google then you’re going to be killed by those labs in the next 6 months. Your users don’t wanna pick models.

@HannahKonnn They just have to somehow pull it off…without the talent, mentality or incentives of a tech company

“AI models trained on public data will become commodities.
AI models trained on proprietary data will become moats.” - Larry Ellison Law firms are sitting on troves of proprietary templates, memos, playbooks, etc.

@willchen500 @TTrimoreau Ok fine, answer changed to mikeoss

@rosieteouk @TTrimoreau Pretty sure they are now the most hated legal tech company. Harvey in second place.

Name a tech company that nobody hates

@Chrisgpt The car wash is portable and can be carried back. Well done Opus

Anthropic why are we failing simple logic on high 🫩

@The_Ai_CPA 💯

This is the greatest time in history to be a professional. Service as Software is the new SaaS and for the first time ever, those of us who actually do the work can build our own software - with little to no input from engineers. I’ve never had this much fun in my career.

@gregisenberg AI slop about AI slop

Every CEO layoff letter in 2026 follows the same template. "Hardest decision I've ever made. AI changed everything. New roles designed for AI-native work. We owe it to our customers. We're choosing to compete." I feel like I'm reading the same letter with different logos

@biglawbro Let’s see what he says when that $10m/y budget gets cut

talked to the GC of one of the biggest, most old school, and yet fastest growing developers in south florida. he has big law experience and is AI aware enough to be using Claude Skills. he isn't thinking in terms of AI for their big deals and doesn't even like what he's seen it do w NDAs. he has a thin legal team and they are hiring a head of AI. his goal is to just not increase their $10m/yr legal spend as they grow.

@w_milczynska Endometriosis and many other chronic conditions impacting women have been massively ignored

chronic health is the rare category where the user has the problem every day for decades, will pay to solve it, and has been actively ignored by tech for 20 years. genuinely don’t understand why more people aren’t building here.

@ZachAbramowitz @biglawbro Smart people also make mistakes. Probably what makes them smart 😉

I've cautioned people who think Harvey/Legora are hype/vapor/wrappers whatever that a lot of smart and very rich people both at lawfirms and VC are betting big on these tools. For the same reason, you probably don't want to bet against Kirkland. They're pretty fucking smart too

@danshipper @every I literally just cancelled my Claude Max this morning in favour of GPT5.5. Hereeee we go again

BREAKING: Anthropic just dropped Opus 4.8—and it is a MONSTER We've been testing for about a week @every and our verdict is they could've just called it Opus 5, it's that good. Here's our vibe check: - Beats GPT-5.5 on Senior Engineer bench. On our toughest benchmark Opus 4.8 scores a 63—a hair higher than GPT-5.5's score of 62, and a full 30 points higher than Opus 4.7. It tackled a ground-up rewrite of a production codebase, and actually built something that works. HOWEVER: Coding performance varied a lot at different reasoning levels. We recommend using it on xhigh for best results. - Incredibly good writer. Opus 4.8 scored a 79.6 on our writing benchmark—measuring models on real-world writing tasks we do all of the time like essay writing, promo email writing, and more. It beats GPT-5.5 by 6 points. It produces well-written prose with fewer "AI-isms". It's also very good at writing in your voice given the right context. HOWEVER: Writing performance also varied with reasoning levels. Medium reasoning had higher incidence of AI-isms—we found best results with high. - Beast at knowledge work. Opus 4.8 is very good at general knowledge work tasks like report creation, research and more. It produced the best PowerPoint one-shot we've ever seen on our deck generation benchmark. - Emotionally intelligent, willing to question the frame. I've also found it to be quite good at talking through psychological or interpersonal issues. It has a high EQ, and it's also good at not glazing and helping to expand your perspective. Its thought process feels extremely rich and dynamic. THE BAD: These days a model is only as good as its harness, and Codex is still a far superior harness to the Claude Desktop app. This has kept me using Codex + GPT-5.5 as my daily driver, but I am flipping back and forth a lot more between Codex and Claude. Anthropic is back baby! Read the rest on @every: every.to/vibe-check/opu…

@loganbrown799 I think there’s enough of a market for Harvey/Legora to remain tech players only but I agree the rise of the AI native last firm is here

It's a matter of time before Harvey/Legora start providing legal services directly and use their tech to take on their current client base. Tech companies and law firms are converging. Law firms are seeing the writing on the wall. Kirkland is developing their own technology. Other law firms are switching to Claude Enterprise to avoid giving Harvey/Legora proprietary knowledge of their workflows. The valuations of Harvey/Legora (11b/5.6b) are surpassing the total addressable market for legal software. The tech world is excited about the services market for a reason. The legal services market is enormous (3trillion) and AI is allowing for it to be disrupted in a meaningful way for the first time. K&E's move here should be a signal to all lawyers about what is going to happen to big law in the future.

@heyitsalexsu You need to label the lady holding up the K&E kid as Claude or OpenAI

Just another day in legal AI

@EverydayAI_ Then it’s worked to make it better for non technical people like me

@rosieteouk the UI is, yes, but the functionality

new codex side panel clean af

@writedotlaw The problem is most private practice lawyers are not incentivised to think about efficiency. In-house lawyers do - we’re often under resourced - so my bet is in-house lawyers will lead adoption

Legal tech used to be something most lawyers could safely ignore. You needed an email. You needed PDFs. You needed to survive the firm’s document-management system. But the machinery underneath? That could stay with IT. In the age of AI, that’s no longer an option. 🧵

@gdb @danshipper Yes Codex is really good for legal work, despite not having a ‘for legal’ or Jude Law marketing campaign

codex is great for any kind of work done with a computer:

@willchen500 Have you tried their word plugins against the Claude or Copilot ones? I just don’t get why anyone would use the Harvey/Legora version of those

The Harvey reddit ama reminds me of a Soviet central committee meeting. Questions submitted beforehand, a friendly press moderator and ChatGPT-generated answers reviewed by PR consultants. The only remarkable part of it was the announcement of 300m ARR. Harvey is winning the distribution game against Legora. Would be very nervous if I were working at Legora. Same costs, same strategy, same product, weaker distribution. Having tried both myself I think Legora actually has a significantly better word-add in that actually works and a slightly better web app. But frankly none of this really matters when you are selling to 50+ year olds who can’t convert a pdf on their own.

@clairevo The good news is they haven’t set up bots to auto reply: “Absolutely! I’ll add that to my dataset. How else may I assist you today?”

PSA if you are a $2 billion company hiring a CPTO with one of the major exec recruiting firms you are paying their recruiters to no-context spam candidates drip campaigns and then ghost them when they reply (i got 2 unpersonalized emails, replied per below, then ignored)

@DanielSmidstrup No you only need the ability to take a vision, break it down into each problem statement, then break those down into the solution building blocks. Then you’re not fully at the mercy of Claude/Codex and can control the outcomes better

Hot take: Vibe coding only works well if you already know how to code.