Just a fun fact - whatever work product you would've delivered in your "10 years in the profession", would be significantly less than the work product generated by Chatgpt in 88 hours to solve a math millenium problem (and just FYI math PHDs usually end up dedicating their entire careers to solve one such math millenium problems).

I get how people on this thread are trying to explain how humans are more capable than AI. I would have been in the same boat if I was new in profession because I would like to bet on my knowledge and capability which in turn will develop a skill which shall be distinct than others.

However, AI has become lot better in the last couple of years and if right context is given, it can give great output in a very short span of time, which humans can't compete. I totally agree that humans can't be replaced completely, but the amount of work a law firm or an in-house department that can do will be more if they use AI. AI is here to stay since its helpful. I am not looking to replace my future hiring plans with AI subscriptions, but I would certainly expect a better work product from team since they have tools at their disposal.

I would suggest young lawyers to spend more time understanding the law and client requirement (by attending more negotiation calls etc.) as that will help you develop your thought process which coupled with technology will make you a better lawyer. Along with that start working on your prompts so that you can give better context to AI. You should be better equipped rather than fight it. I also understand that young lawyers don't have luxury of time due to immense work load, but threat is knocking on our doors, so for the sake of self-preservation see what one can do.

PS: I have spent 10 years in the profession.

Did some calculations on the above.

The agent swarm generated an output of 130 billion tokens. 130 billion tokens will roughly be around 97.5 billion words.

So let's split 97.5 billion words between say 10,000 human researchers. So each researcher would need to write/produce at least 9.75 million words. 9.75 million words comes out to around 130 books (each book containing 300 pages and each page containing 250 words per page).

So 10,000 human researchers would need to write 130 books, each to match the output generated by OpenAI in 88 hours. Look at it from a different POV - that's 13 lakh books (10,000 * 130) worth of content!

These numbers are insane!

11,000 agents (or bots in layman terms) used a shit ton of compute for 2-3 whole days to solve the NS problem. I read some AI founder saying that the amount of work that these agents did, in human labour terms, was equivalent to 100 years of human labour output. So like if 11,000 human math researchers worked for 24 hours for 100 years, only then would the human output would be comparable to the output generated by these agents.

I read this and I was like even if it's 99% exaggerated and like the comparable human output is 1 year, even then this news is freaking impressive. 1 year or possibly many more years worth of output in 2-3 days. If this true then this is nothing short of the AI that we've been seeing / reading in sci fi movies and stories.

I've been overthinking on how this affects the billable hours model ever since

I quit my job at a law firm at 10 months PQE and I’ve been job hunting for the last 6 months. I really wanted to stick to M&A, PE/VC as it’s always been my practice area but now I don’t know if it’s even possible. It feels like no one is hiring freshers anymore and I might have to switch practice areas.

the same way the Radiologists shat themselves when the ai doom trollers pretended like they achieved something and will displace jobs.

They very much stole navier stokes as well, the team claims chatgpt stole unpublished data and there have been multiple allegations of shady behaviour from those clowns.

Someone good enough to solve complex math would not need to pretend that their AI is hacking the pentagon for clout. AI is incredible, but AI taking my job is some way away and this isnt the time to panic because by then I'd be in a position to not care about money.

yes, AI still cant reason, if you think a lawyers or any profession really is limited to following rules to the T and not trying to apply it in a certain manner, you should not be a lawyer at all.

math as a skill set will die, anything where training is purely on the basis of following rules will die out, luckily thats almost no profession in its entiretiy.

Read this guys: https://rahulmatthan.com/ex-machina/2026/494.-the-journey-is-the-point/
Just like how tech companies have started hiring back their developers after realising that AI is more expensive compared to human talent, I am sure the legal industry will also have this realisation.

Person who posted the comment here.

Funnily enough, the founding partner of my firm just further echoed this and stated exactly what most people knew for ages on linkedin.

you can replace an a0, but you cant replace an SA.

problem is you cant get an SA without an A0 so A0s wont ever go away.

The only competitors to OpenAI, Anthropic, and Grok are chinese open source models (Kimi, Qwen, Deepseek etc.). Whatever regulations the US Government creates to slow down AI development will only affect them and not their chinese competition. Wouldn't asking for a state mandated slow-down actually give their Chinese competition a leg up instead of themselves?

Brother the humans, Buckmaster and Apologe, solved a different problem. They solved the force variant of the Euler Equations rather than the Navier-Stokes that OpenAI solved. The Euler equations are a simpler, idealised stepping stone (without viscosity) compared to the full Navier-Stokes equations. So even if we give the humans the benefit of the doubt, AI still took the solution of a simplified variant to a millennium prize winning solution. That's a meaningful contribution. The ethics of that are debatable because OpenAI didn't credit anyone, but I'm not arguing about that here.

Also a few days after this happened, 26 field medalists (equivalent to the Nobel Prize for mathematics) signed a letter saying how powerful (and mismanaged) AI use in Maths is. https://mathandai.org/

First line: "Over the last few months, the mathematical capabilities of LLMs have improved dramatically, to the point that they can solve major outstanding problems in many fields of mathematics."

These words aren't by AI doom trollers, they're literally by the leaders of the math research community. You still think AI is nothing serious to be worried about?

So you mean to say that a computer algorithm woth zero legal knowledge will produce a far better work than someone with 10 year experience ? Then let's just scrap law as a profession and let AI take over arguments also?

AI is making people lazy. I see people at my first using it for basic take like drafting an email just shows the level of underconfidence.

Unethical to charge your client if your draft is nothing but a humanized version of an AI output.

Work is not infinite because if its reptitive or a model can be trained on it, large companies or in house teams can get it done via ai. The only factor remains is executives want to shirk accountability which is where law firms will play a role in the future. Lawyers will have to work 2x to make the same amount, number of entry level jobs will diminish significantly

You are right to be worried. As someone at a senior level, running a firm that is working with AI to improve - I can tell you that AI as a whole is significantly going to change the landscape for the work to be done.

Just the other day, I was able to put together the first draft of an entire complex agreement using AI. I had to burn about 15K worth of credits, but it was well worth it in that the quality of output was very much worth paying 15K for - and it did that job in about 4 hours - while even a very good PA / SA would have taken atleast a day with no guarantee of quality.

So what does this mean for you and anyone else who is starting out in the profession:
(a) Judgment / understanding context beyond the written text / being able to think deeper and more comprehensively - those will be skills people will look for in the future. So focus on understanding commercial context, understanding different approaches to solving complex problems, understanding why something is the way it is. In my time, you had no way of accessing this knowledge or even knowing what to ask. Today AI can take you quite deep and act as a thought partner.

(b) A natural caveat to that is that lots of juniors and lawyers in training are using AI as a substitute for thinking and understanding, simply because its easy to get answers from AI and then repeat it. This will be the biggest cause of down-fall for young folks coming through. If I ask you for your thought on something and you just repeat what AI told you without understanding (and trust me - we know), then why do I need you - I can just train an AI to respond to me in those situations. So that is what you will need to avoid.

All of the above, notwithstanding, there will always be space for human lawyers - its just that drafting has gotten a lot easier - so I guess that's something. But again - like (a) and (b) above, you have to draft ground-up before you can use AI to do the drafting - so keep in mind as well.

Hope this helps. Happy to answer more questions.

The nature of the job will change, associates won't be needed to spend as much time, so that's good. However, learning will also be a lot lesser - people tend to stop using their brains and rely on AI too much sometimes, that's the danger.

Work is infinite, so I don't think hiring will be a problem.

bhai hinton who ran google deepmind and created the model behind all current AI said that radiologists would be gone in 5 years entirely, 11 years ago, guess what, we are the biggest crunch for radiologists ever to the point where surgeons take it up part time for more money.

Listening to AI doom trollers cost many people a guranteed job in the US earning 6 crores a year on average.

88 hours it solved a mathematics problem? they stole the answer and then blackmailed the HUMANs who actually solved it.

AI will kill mind numbing work, now the sad reality is most corporate jobs were mind numbing work, if you were possessing the skill set of having sophisticated language while doing nothing more challenging than a white collar employee, you will get swallowed alive, if you have a brain that functions, you are good.

the recent doom troll about AI killing humans was something these AI bros were spewing in the 1980s saying its a few years away since then, they will say it in 2070s too and your grandchildren will be hearing the same nonsense, its just IPO hype.

The twitter post that kicked this off came 7 minutes after an article from WSJ quoting the guy leaving anthropic, he said the CEOs downplay the AI risk? Bruh those AI bimbos have been going on and on about how AI will kill all of us for ages, this year both dario and sam have said in podcasts how AI will kill all humans but they're downplaying risk.

everyone in AI is incentivised to make AI seem better cause all their employees got paid in stock options which only have value if their ipo next month is going to go well.

Further another aschenbrenner who runs a famous trending doom troll and then decides to start a company with billions in funding is also likely in fact most probable.

Look at guys who have left big AI wings like Yann Lecun who won a turing award btw, who says AI cant compete with cats let alone humans in terms of reasoning and rationale.

As you all may know by now, an internal ChatGPT model has managed to solve a millennium problem, the hardest set of problems in mathematics, in 88 hours. We also have Sam Altman and Dario Amodei, the CEOs of the ChatGPT and Anthropic, calling for government regulation because they believe AI is too powerful to be regulated privately. Think about that for a sec. Private, capitalist companies calling for government intervention. That itself is a scary thought. I'm a 2nd year at a T1 NLU and I'm genuinely concerned about the future of this profession if AI continues to evolve the way it has been evolving. What are yalls opinions and do you think AI genuinely will be a threat to the associate/fresher jobs in the legal industry in ~4 years from now?