It’s probably a mixture of things including direct control over how the api is called and used as pointed out above and giving a discount for using their ecosystem. They are in fact a business so it should not surprise anyone they act as one.
It might well be a mixture, but 95% of that mixture is vendor lock in. Same reason they don't support AGENTS.md, they want to add friction in switching.
The problem is even if an OSS had the resources (massive data centers the size of NYC packed with top end custom GPU kits) to produce the weights, you need enormous VRAM laden farms of GPUs to do inference on a model like Opus 4.6. Unless the very math of frontier LLMs changes, don’t expect frontier OSS on par to be practical.
I feel like you're overstating the resources required by a couple orders of magnitude. You do need a GPU farm to do training, but probably only $100M, maybe $1B of GPUs. And yes, that's a lot of GPUs, but they will fit in a single datacenter, and even in dollar terms, there are many individual buildings in NYC that are cheaper.
I refer you to the data centers under construction roughly the size of Manhattan to do next generation model training. Granted they’re also to house inference, but my statement wasn’t hyperbole, it’s based on actual reality. To accommodate the next generation of frontier training it’s infeasible for any but the most wealthy organizations on earth to participate. OSS weights are toys. (Mind you i like toys)
There's already an ecosystem of essentially undifferentiated infrastructure providers that sell cheap inference of open weights models that have pretty tight margins.
If the open weights models are good, there are people looking to sell commodity access to it, much like a cloud provider selling you compute.
I up you to continuous reporting. Audit should be inherent to the system, not a process after the fact. As a public company all owners should have access to daily closed books, and all companies should be able to close their books daily in 2026.
Every six months being the cadence we learn how our companies we own are doing is absurd. It leads to really long dark periods. Also for employees it means we can only divest in a semi annual window. Our carry risk is extensive and expanding.
This is about hiding truth longer, which is the MO of this administration top to bottom.
That is an absurd cadence. It is extremely expensive to do this reporting; an an enormous amount of useless activity is slaved to providing it in companies that need to. This is literally a call for more bureaucracy theater.
The obvious net effect is that companies would structure themselves to no longer have the reporting requirement, as the cost of reporting exceeds the benefits. That would not benefit society at large.
The reason quarters take so long to close is because the numbers are being fiddled with. There's no reason someone shouldn't be able to close a quarter and report the numbers with the automation we have today in technology, meaning without some magic AI/LLM, other than people are constantly trying to reclassify expenses or income in a way that saves the quarter
Why, after 30-40 years of modern computing in accounting does it still take a month to close the books? I worked at a public company that was $100m revenue yearly and it took a whole month to close the books. Absolute insanity. Even AT&T or Verizon or GM should be able to report at least weekly.
This is a naive view of what reporting entails and the difficulty of coalescing a report that meets the requirements of the audience the report is for. It isn't a numbers dump from a database, it requires substantial interpretation of things that the database does not and cannot contain. It isn't fiddling with the numbers, it is that the numbers can't contain things relevant to their representation for external parties as a legal matter.
When I have been in positions where reporting was a necessary part of my job, reporting related activity probably consumed 1/3 of my time. Even in highly optimized contexts, it consumes a stupid amount of time and the impact on the consumers of those reports is often quite low. It is almost a total waste of time.
There should be some reporting but the current cadence and requirements is way too high for many large companies. Reporting doesn't have infinite ROI.
> it requires substantial interpretation of things that the database does not and cannot contain.
Do you have examples? This seems like something that is a solvable problem, and from the outside it can seem like it is only about not being willing to switch to a new paradigm. That unwilling ness can come from avoiding real consequences like loosing a competitive edge due to allocation of resources to the switchover.
When people think of automation I'm assuming their thinking of the financial statements (balance sheet, income, cash flows, equity).
Reporting also contains narrative explanations by management of: the company's financial health, updates on any new or existing market risks and the company's strategy to deal with them, any changes to controls or accounting procedures, updates on any new or existing litigation, and more.
These reports need to be certified for truth by the CEO, CFO, and relevant officers under penalty of 10+ years in jail and millions of dollars in fines personally.
It's also common to do a press release, earnings call, and investor presentation but those aren't required.
I meant just closing the financial records, not coming up with the shareholder marketing. It can take a month just to find out if you "made" the quarter or not, mostly because accounting and finance is combing through every line item to see if they can recategorize it in a way that makes the numbers look better but doesn't result in them going to jail
In what should be a very black and white line of work there is a ton of judgement and negotiation involved
Do you understand that as a legal matter these must be good faith representations of the current state to the best of your knowledge? You can’t serve up intentionally stale information without inviting legal repercussions. The preparation process takes weeks. This is a very serious legal matter.
These are being revised and updated right up until the point they are released to provide the most accurate reporting possible.
You gravely underestimate the legal seriousness of these reports.
So let's try to think of solutions instead of giving up. A law that requires daily disclosure can change how the reporting works so you don't need to update those category decisions 200 times.
Does the technology already exist? Things are almost never only a tech issue alone. That does not mean tech can not help, even if the tech that would help is currently impractical. What is impractical now though may not be in 10, 20, 50 years.
Going over what I quoted:
> You can’t serve up intentionally stale information without inviting legal repercussions.
Keeping information fresh and up to date is something technology has helped with in many areas. If there is a reasons why it can not help here then I an interested in why or that the current tech already does a good enough job in this area.
> These are being revised and updated right up until the point they are released to provide the most accurate reporting possible.
Technology can help verify last minute changes, running a test suite for example or similar. How hard that is to make or maintain though may make impractical.
> You gravely underestimate the legal seriousness of these reports.
Having an audit trail and known processes may be helpful here too if the current tooling is not adequate.
I quoted parts of the comment that looked like areas where tech has already helped in other areas. What I want to find out are details about what exists, why people think it can not be better, or why pervious attempts have failed, or why things are currently optimal.
My fiancee is the accounting manager at a university. Why? Because people don't submit expenses on time, invoices are delayed or some still done manually, and all manner of things. Even for them it can take a couple of weeks.
While there may be some "hijinks" (in their case, institutional advancement likes to steadily rearrange endowments or donations to take advantage of offers to match donations, etc., but that's not really a delay, as accounting basically says things like "No, that gift has already been spent"). Even with things like Concur or Expensify, expenses aren't classified on time, submitted for reimbursement, etc.
It’s only expensive because it happens so episodically that it doesn’t require automation. Automation leads to scale leads to reduction in cost. The analysis humans do on top of it can be done through a periodic filing, but the totality of disclosures can be done continuously other than the periodic human opinion fluff. The notes and details can be filed as they are relevant without undue burden. (I.e., a large onetime expense can be explained as it happens - I assure you it is being explained internally at that time in more detail).
I was at a large Wall Street firm which closed its books daily and has done for decades. They disclose daily to the fed and others. It was work for sure but the benefits of constantly knowing your business far out weighed the cost. So I don’t buy that it turns into more theater, you can’t do theater at a continuous pace. Theatre takes time, and the level of theatre increases as the pace of disclosure decreases.
You don't have all the relevant invoices etc at on time. Some of that takes quite awhile. Especially inter country purchases and sale transaction information.
This doesn’t get better when you have a quarterly or semiannual deadline. It’s just the scale might be smaller. However you would handle them in the same way and either disclose on an accrual basis or on a cash basis, but either way, you do it as you know it.
I’ve been using 1M for a while and it defers it and makes it worse almost when it happens. Compacting a context that big loses a ton of fidelity. But I’ve taken to just editing the context instead (double esc). I also am planning to build an agent to slice the session logs up into contextually useful and useless discarding the useless and keeping things high fidelity that way. (I.e., carve up with a script the jsonl and have subagent haiku return the relevant parts and reconstructing the jsonl)
double escape gets you to a rewind. not sure about much else.
the conversation history is a linked list, so you can screw with it, with some care.
I spend this afternoon building an MCP do break the conversation up into topics, then suggest some that aren't useful but are taking up a bunch of context to remove (eg iterations through build/edit just needs the end result)
its gonna take a while before I'm confident its worth sharing
Yeah just selective rewind. Selective edit where you elide large token sinks of coding and banging its head on the wall is what u mean. Not something I’ve seen done yet but there’s no reason - I suspect if you do a token use distribution in programming session most goes to pretty low semantic value malarkey.
The problem is humans are really bad at perceiving externalities at this scale, cause / effect between small actions and large effects, and effects that play out over the span of their lifetime rather than the span of their day. The denialist rationale shifted over the years from doubting the very basis of the science, to claiming its just a short term blip, to its natural long term cycles, to … everything that involves not looking up.
I think the truth is we won’t really take this seriously globally until the changes are so severe that it’ll take generations to undo if ever.
My employer is pretty advanced in its use of these tools for development and it’s absolutely accelerated everything we do to the point we are exhausting roadmaps for six months in a few weeks. However I think very few companies are operating like this yet. It takes time for tools and techniques to make it out and Claude code alone isn’t enough. They are basically planning to let go of most of the product managers and Eng managers, and I expect they’re measuring who is using the AI tools most effectively and everyone else will be let go, likely before years end. Unlike prior iterations I saw at Salesforce this time I am convinced they’re actually going to do it and pull it off. This is the biggest change I’ve seen in my 35 year career, and I have to say I’m pretty excited to be going through it even though the collateral damage will be immense to peoples lives. I plan to retire after this as well, I think this part is sort of interesting but I can see clearly what comes next is not.
Why are you excited for this? They’re not going to give YOU those peoples’ salaries. You will get none of it. In fact, it will drag your salary through the floor because of all the available talent.
I’m excited as a computer scientist to see it happening in my life time. I am not excited for the consequences once it’s played out. Hence my comment about retiring, and empathy for everyone who is still around once I do. I never got into this for the money - when I started engineers made about as much as accountants. It’s only post 1997 or so that it became “cool” and well paid. I am doing this because I love technology and what it can do and the science of computing. So in that regard it’s an amazing time to be here. But I am also sad to see the black box cover the beauty of it all.
I'm very confused about this. Salary is only one portion of your total compensation. The vast majority of tech companies offer equity in a company. The two ways to increase the FMV of your equity is: increase your equity stake or increase the value of the total equity available. Hitting the same goals with fewer people means your run rate is lower, which increases the value of your equity (the FMV prices in lower COGS for the same revenue.) Also, keeping on staff often means you want to offer them increased equity stakes as an employment package. Letting staff go means more of that available equity pool is available to distribute to remaining employees.
We aren't fungible workers in a low skill industry. And if you find yourself working in a tech company without equity: just don't, leave. Either find a new tech company or do something else altogether.
Equity is negotiable just like salary, and if supply of developer labor increases with the same or less demand, you'll get less equity just like you will get less salary.
I can't believe the person you replied to thinks that they're going to get some magical more amount of equity because you can hopefully do more with fewer people. That's assuming the entire business landscape doesn't also change with AI, disincentivizing so much investment in companies in the first place because someone else with AI can create a competitor in a shorter amount of time...
In the last three startups I worked at I didn’t bother exercising my vested equity - even a successful exit would at best triple the price of those shares - not worth the risk. One of those three startups already failed.
5.4 is the one fine tuned for autonomous mass murder, automated surveillance state, and money grabs at any cost. It’s really hard to lump that into the others as it’s a fairly unique and specialized feature set. You can’t really call it that tho so they have to use the numbers.
I’m pretty glad I’m out of the OpenAI ecosystem in all seriousness. It is genuinely a mess. This marketing page is also just literally all over the place and could probably be about 20% of its size.
This is a huge step over M4 32GB 153GB/s memory transfer
For local LLM this make it a replacement for a DGX Spark, which offers a third of the transfer speed and is not something you toss in your backpack as your laptop. It’s practically useful for a lot of local use cases and that I think is the 4x factor (memory xfer) - but the 128Gb unified headroom tremendously improves the models you can run and training you can do.
What is truly amazing is the M1 Max is 400GB/s. 5 years later and we still only hit 1.5x on memory bandwidth. It's quite fascinating how high Apple spec'd it back then with apparently little foreknowledge of how important memory bandwidth would become, and then conversely how little they've managed to improve it now when it's so obvious how important it is.
The reason for that is that most memory bandwidth bumps come with new memory generations. For example an early DDR4 platform (e.g. Intel Skylake/Core iX-6000) and a late one (e.g. AMD Zen3/Ryzen 5000) only differ by 1.5x as well, typically.
The same trend is visible in GPUs: for example, my RTX 2070 (GDDR6) has the same memory bandwidth as a 3070 and only a little bit less than a 4070 (GDDR6X). However, a 5070 does get significantly more bandwidth due to the jump to GDDR7. Lower-end cards like the 4060 even stuck to GDDR6, which gave them a bandwidth deficit compared to a 3060 due to the narrower memory buses on the 40 series.
I run Qwen 3.5 30B MOE and it’s reasonable at most tasks I would use a local model for - including summarizing things. For instance I auto update all my toolchains automatically in the background when I log in and when finished I use my local model to summarize everything updated and any errors or issues on the next prompt rendering. It’s quite nice b/c everything stay updated, I know whats been updated, and I am immediately aware of issues. I also use it for a variety of “auto correct” tasks, “give me the command for,” summarize the man page and explain X, and a bunch of tasks that I would rather not copy and paste etc.
Each of those clauses have a DoD policy carve out as an exception which says basically they can do whatever they want if they want to do it, but won’t be able to if they don’t want to do it.
reply