The differentiating factor will be access to proprietary training data. Everyone can scrape the public web and use that to train an LLM. The frontier companies are spending a fortune to buy exclusive licenses to private data sources, and even hiring expert humans specifically to create new training data on priority topics.
> At what point are the models going to all be "good enough", with the differentiating factor being everything else, other than model ranking?
It's already come for vast swathes of industries.
Most organizations have already been able to operationalize what are essentially GPT4 and GPT5 wrappers for standard enterprise usecases such as network security (eg. Horizon3) and internal knowledge discovery and synthesis (eg. GleanAI back in 2024-25).
I agree, and most of my peers do as well. This is why most of us shifted to funding AI Applications startups back in 2023-24. Most of these players are still in stealth or aren't household names, but neither are ServiceNow, Salesforce, Palo Alto Networks, Wiz, or Snowflake.
Foundation Models have reached a relative plateau and much of the recent hype wasn't due to enhanced model performance but smart packaging on top of existing capabilities to solve business outcomes (eg. OpenClaw, Antheopic's business suite, etc).
Most foundation model rounds are essentially growth equity rounds (not venture capital) to finance infra/DC buildouts to scale out delivery or custom ASICs to enhance operating margins.
This isn't a bad thing - it means AI in the colloquial definition has matured to the point that it has become reality.
I thought companies can't profit off of unclaimed/abandoned money.
Was under the impression that funds like that eventually get handed over to whatever state agency is responsible for dealing with unclaimed property.
(If so, the cause might just be incompetence rather than greed or malice - not that incompetence is any better than malice when it comes to handling people's money)
its surprisingly easy to get away with murder (literally and figuratively) without piercing the corporate veil if you understand the rules of the game. Running decisions through a good law firm also “helps” a lot.
A bit over five years ago, someone struck and killed my friend in a crosswalk. He was a fellow PhD student. It was on a road with a 30mph limit but where people regularly speed to 50+mph.
He was an international student from Vietnam. His family woke up one day, got a phone call, and learned he was killed. I guess there was nobody to press charges.
She never faced any accountability for the 'accident'. She gets to live her life, and she now runs a puppetry education for children. Her name even seems to have been scrubbed from most of the articles about her killing my friend.
So, I think about this regularly.
I was a cyclist at the time so I was aware of how common this injustice was, but that was the first time it hit so close to home. I moved into a large city and every cyclist I've met here (every!) has been hit by a car, and the car driver effectively got only a slap on the wrist. It's just so common.
> Her name even seems to have been scrubbed from most of the articles about her killing my friend.
I'm somewhat surprised there were even articles? Are road fatalities uncommon enough in the US that everyone gets written up? Or was this a special enough one?
Not sure if this is true for every university, but when someone in the community dies, especially a student, there's usually at least an article about it.
It’s funny to think that, like AI, people take actions and use corporations as a shield (legal shield, personal reputation shield, personal liability shield).
Adding AI to the mix doesn’t really change anything, other than increasing the layers of abstraction away from negative things corporations do to the people pulling the strings.
Allowing student debt to be canceled during bankruptcy would be a good first step (possibly even better than canceling student debt across the board).
To your point, making it easy to cancel debt teaches borrowers that debt isn’t a serious thing.
Requiring someone go through bankruptcy (and all of the associated negatives on your credit score, etc) seems like a good tradeoff. Allows you to get out from under the debt (the entire purpose of bankruptcy in the first place..) while not letting everyone pretend the debt never existed (need to live with the impact of bankruptcy on your ability to borrow in the future)
I don’t know why we don’t hear more people lobbying for this. I guess it’s because the sound bite isn’t as sexy.
Bankruptcy affects your credit score for 7-10 years. Someone who graduates from college in their early 20s with six figures in debt could file for bankruptcy immediately and have it be off their credit history by the time they've saved up a down payment and want to get a mortgage.
There is also the obvious drawback that if more people can discharge the debt, the interest rate goes up, and then everyone else has to pay for the people who took out loans they didn't pay back.
> Someone who graduates from college in their early 20s with six figures in debt could file for bankruptcy immediately and have it be off their credit history by the time they've saved up a down payment and want to get a mortgage.
So change the bankruptcy law? It’s a pretty easy fix. Create a whole new chapter if that’s what it takes. Make it dischargeable only after 7 years of nonpayment, do means testing… bankruptcy law already has these kinds of nuances built in.
> Make it dischargeable only after 7 years of nonpayment
You don't really want to give people an incentive for nonpayment.
> do means testing
Bankruptcy already does that. But what are your "means" the day you graduate from college and haven't yet found a job, or temporarily take a low-paying one on purpose to meet the eligibility requirements?
You would need something like, deferred payments while you're unemployed but if you subsequently find a job then you have to pay, instead of one-time permanently discharging the loans. Except that's how it already works.
> The law is not meant to cover every possible permutation of circumstances and motivations.
This isn't a permutations issue. We know the specific shape of the problem: 18 year olds from poor and lower middle class families don't typically have existing assets with which to secure a loan, so if they can't secure it with their future earnings, they can't get one, and then they can't afford to go to college.
> If student loans are dischargeable in bankruptcy, lenders will price it in or refuse to lend without a gurantor.
That's the problem. The inability to discharge them allows the borrower to get a much lower interest rate than they otherwise could for unsecured debt issued to someone with minimal credit history, or find someone willing to loan them the money to begin with.
It was set up this way so that people could go to college.
Imagine a world where lenders charged different interest rates depending on the risk profile of each school.
Lower interest rates for schools where graduates repay their debt, higher interest for schools where many people default.
Assuming it wouldn’t disproportionately affect disadvantaged populations, that could be an interesting way to incentivize schools to get their shit together and prepare students for starting their career
Don't have a dollar amount that you repay. Rather, your student loan payment is x% of (your income minus the average rate for those with a high school diploma) for y years. Forgiveness programs for certain fields go away--instead, the tab gets picked up perhaps with a multiplier. Disability, death? Irrelevant--a dead person generally makes nothing, the amount owed is $0. (Generally makes nothing because there can be ongoing income from something they produced. That would be subject to the loan repayment.)
Devils advocate: I don’t think this would work due to the cash flow uncertainty the models causes.
If universities don’t know how much they’re going to bring in over the next few years, they won’t be able to budget effectively.
And then there’s the question of whether it’s acceptable for the lender to collect more than what was borrowed. E.g. if I graduate college, start a company, and sell it for $100 million.. am I then paying my alma mater (or lender) millions? If so, would universities make more money from the commons or are they banking on a very small percent landing extremely lucrative gigs post-graduation? I don’t think we want the model to resemble startup financing, where nearly all fail and a small handful pay for the rest (that works for startups, doesn’t work for people’s careers)
I like the concept though. In 2010’s when I was entering college, I actually made a website trying to solicit someone to pay for my education in exchange for a percentage of my future earnings. I found no takers at the time.
That effect would be drowned out by all the people defaulting en masse because getting out of a six figure unsecured debt is worth more than a temporary hit to your credit score.
If that’s true, why isn’t everyone already doing it? Especially if 87% of student loans are forgiven during bankruptcy - maybe people just aren’t aware it’s possible.
"Possible" and "easy" are two different things. 87% is of the people who applied for it, but they would be the ones most likely to have it granted because there is little reason to pay lawyers to file a form which is likely to be rejected. But the more you relax the requirements, the more people do it, which is of course the problem.
People who start lobbying for it quickly discover that 87% of people who petition for their student debt to be cancelled in bankruptcy get it (https://www.cnbc.com/2025/12/29/bankruptcy-student-loan-borr...). I support removing the special treatment entirely, but ultimately most student debt holders don’t go bankrupt.
> Still, few people pursue the option because of a “pervasive” myth that the loans can’t be included in the proceeding
It also says nothing about whether the person actually goes bankrupt, just which debts are discharged, which is one of the key parts of the bankruptcy process. Certain debts are discharged because the person can’t pay them back, which is the point of going into bankruptcy court.
IMO it’s (unfortunately) the public’s responsibility to learn the lesson that LLM’s shouldn’t be trusted without double checking the source — same position Wikipedia was in 10 years ago. “Don’t use Wikipedia because it has incorrect information” used to be a major concern, but that seems to have faded away now that Wikipedia has found its place and people understand how to use it. I think a similar thing will happen with LLM’s.
That opinion does not take the responsibility away from LLMs to continue working on educating people and reducing hallucinations. I like to think of it as equal responsibility between the LLM provider and user. Like driving a car - the most advanced safety system won’t prevent a bad driver from crashing.
We also are working on crowdsourcing methods, but it's hard because almost everyone involved in the development of this project is a volunteer that either works for a company already or is a startup founder (me)... so is very tricky to find time.
I used it for about a year with a small team. It worked well for what it does, but the functionality is definitely stripped down and barebones compared to Slack. I don't remember any performance or reliability issues.
Hard disagree. We use both in my company. Google Chat is definitely better than Teams for actual collaboration: it's easier to track unread messages in "Home" (it's the "inbox"), and channels (called "spaces") are much better designed (they are conceptually closer to Slack's channels). Also, it's not crashing all the time. What's missing: the message editor doesn't support nested bulleted lists, we can't archive a space/channel.
Your comment matches my experience closer than the OP.
A link disappearing isn’t a major issue. Not something I’d worry about (but yea might show up as a finding on the SOC 2 report, although I wouldn’t be surprised if many auditors wouldn’t notice - it’s not like they’re checking every link)
I’m also confused why the OP is saying they’re linking to public documents on the public internet. Across the board, security orgs don’t like to randomly publish their internal docs publicly. Those typically stay in your intranet (or Google Drive, etc).
> although I wouldn’t be surprised if many auditors wouldn’t notice
lol seriously, this is like... at least 50% of the time how it would play out, and I think the other 49% it would be "ah sorry, I'll grab that and email it over" and maybe 1% of the time it's a finding.
It just doesn't match anything. And if it were FEDRAMP, well holy shit, a URL was never acceptable anyways.
Is it?
At what point are the models going to all be "good enough", with the differentiating factor being everything else, other than model ranking?
That day will come. Not everyone needs a Ferrari.
Edit: I misread the parent, I think they're saying the same thing.
reply