> I understand that Microsoft is planning to remove most of its Claude Code licenses and push many of its developers to use Copilot CLI instead. While Claude Code has been a popular addition, it has also undermined Microsoft’s new GitHub Copilot CLI coding tool — a command line version of GitHub Copilot that runs outside of development apps like Visual Studio Code.
There's definitely a way to use Claude code that is token conscious.
I've tried throwing unsupervised agentic software factory workflows against the wall, and they burned through my tokens like nobody's business but didn't produce much.
Supervised, human-in-the-loop process on the other hand is much more productive but doesn't consume nearly as much. Maybe that's why everyone's pushing agentic approaches so much.
The current thinking is automated agents is what turns this from an industry in the tens of billions to a multi trillion dollar one. So yes you are right on the money, agents stimulate demand for this thing they've built.
So then it's kind of like the government. The more incompetent and inefficient it is, the solution is always paying more taxes to hire more of the same incompetent and inefficient government.
There is always a quantity of lubricant that can get any machine moving. Just add so much that you create an all consuming river of lube and watch your thing sail away.
At the enterprise level though, its going to be hard to want to use a service in which costs are not predictable, and keeping those costs under control requires employee training.
To be fair, the cost of software development has always been fairly unpredictable. What may be different is that the cost used to be roughly proportional to man-hours spent, while now the number of agents running in parallel may be less predictable.
The cost per month is 100% known and always has been. What has been variable is the rate of delivery. AI is different and can be substantial in countries with lower wages.
> To be fair, the cost of software development has always been fairly unpredictable.
Yes, but in a "oops this is gonna take another two months to finish" kind of way, not the "oops this is the 12th time this month 8 developers have burned $2K in tokens in a single day and no one really knows how it happened" kind of way.
since those headlines started ive felt it just encouraged inefficiency. "say as much as you can without saying anything." if you were accomplishing your task the need for more would end, thus there is incentive to never succeed.
> There's definitely a way to use Claude code that is token conscious.
Colleague used Sonnet 4.6 on some pretty normal agentic coding tasks through AWS Bedrock to keep the data in the EU, 100 EUR usage in a single day. In comparison, the Mistral subscription costs about 20 EUR per month and we tested that for similar tasks it was okay, the usage got to around 10% of that monthly limit in a single day. Or Anthropic's own Max (5x) plan where you get way, way more tokens to do with as you please.
I feel like the sweet spot is having a monthly subscription with any of the providers (you're subsidized a bunch), but if you have to pay per tokens, now I'd just look in the direction of what tasks DeepSeek would be okay for, sadly probably not in the situation above. For a startup, though...
On the other hand, this feels a bit hypocritical:
> It was part of an effort to get project managers, designers, and other employees to experiment with coding for the first time, and sources tell me that Claude Code has proved very popular inside Microsoft over the past six months.
They're gonna say that the future is all AI... until they get the bill.
I was trying to get a better sense of the time cost quality matrix of these, so I threw together a quick eval of Sonnet 4.6, Mistral's dev model, and Opus 4.7 (figuring it's what you'd use if you were on Max).
The results for a function implementation and test of levenshtein distance in js are pretty similar but Mistral is 30x cheaper than Opus 4.7 and 4x faster than Sonnet 4.6.
The one detail I did forget to mention is that if anyone goes with the Mistral subscription (instead of paying per-token), then the Mistral Vibe tool gives you their Medium 3.5 model by default, with a 200k token context. It will probably be enough for plenty of tasks, though there's also a noticeable difference between that and up to 1M.
98.6% cache hits doesn't distinguish an efficient workflow from an overly chatty linear agent repeatedly reusing the same context. Plus, it says nothing directly that the process has good useful progress per token.
You pay for cache hits on every turn and even with the newest architectures longer context is slower/more energy intensive. Constructing concise turns that reuse prefix and stop when the new context is no longer useful help, as does pushing generation down into cheaper models while using stronger models for verification.
My experience as well... I've only hit Antrhopic's 5hr threshold a few times, and two of them was within a half hour of the window. Also, all three times I'd already accomplished a LOT.
I tend to work with the agent, and observe what's going on as well as review/test and work through results/changes. I spend a lot more time planning tasks/features than the execution, even using the agent as part of planning and pre-documentation. It works really well. I don't think people burning through the 5hr allotment in under an hour are actually reviewing/QC/QA the results of what they're doing in any meaningful way, and likely producing as much garbage as good (slop).
I'm really curious as to HOW the MS employees were using the agents as much as what they were doing.
I suspect subscription limits are quite a bit higher than the equivalent tokens their dollar cost could purchase. I similarly feel like I can get a lot done with a $20/mo Claude Pro subscriptions, but also can easily spend $10-20/day at API pricing with similar usage.
Personally I prefer the API pricing because I feel like I'm not going to get rug pulled on my work. When it comes to personal stuff, I use the shit out of my sub, but it's not making me money.
I've launched an internal demo of Claude Code and Deepseek on the same day and we burned through our monthly allowance for Claude in just over a week, with more than a half of that budget being spent in one day. With DS people are unable to go through that same amount of money in a month, not even close.
With that Claude feels like an expensive toy, while DS is a shovel, purely because developers do not feel like they are eating into a precious resource while using it. Also it does not feel like there is much of a difference in capability between Claude and DS-pro. DS-pro and flash do feel like sonnet/opus and haiku, but flash is still very-very capable.
After 2 weeks of Claude getting progressively worse and worse, today was the final straw.
I don't care if they have a phone app. The model is COMPLETE garbage after you subscribe long enough and they think they've "got you".
I can't code on my phone if the model literally moves in the wrong direction and does the opposite of what I tell it to. If I wanted to make my code worse, I'd just randomly commit garbage. I don't need a mobile app for that.
Gemini got a big reduction in usage limits this week. There was backlash and they added 3x usage for Antigravity a day later but I haven't really tried it out to get a feel for it yet.
Thus does kind of beg the question: If developers are being laid off because AI is better/faster/cheaper or makes all their people 10x or whatever fig leaf, what happens if the required tooling ends up being more expensive? From the investor’s point of view is the drag of employee costs better or worse than a ballooning expense item?
I suppose if it all works out it'll end up way more expensive than the employees the models displaced ever were. These kinds of technologies usually end up as an oligopoly at best, and those players will have a wide moat by then, and the things these models build will be tweaked such that no other model or human being can realistically work on them anymore, and then they can price gouge everyone to the brink of unprofitability.
The model provider would be like a union, at least if unions had absolute control over their members, could take them all away at any time forever with no substantial negative consequences to itself, and spend billions on employer lock-in so switching to the competition is worse than paying the 12% model salary raise.
I suspect AI would have to get drastically more expensive before it starts looking worse than payroll. If one developer using Claude Code can effectively substitute for 2 developers, you are already coming out ahead at current API pricing assuming very heavy usage, your cost is going to be ~1.5x developer (factoring in beyond salary - benefits, PTO, the other overhead that comes with having employees).
So you're getting 2 for the price of 1.5. Scale that up to 500 devs at a big company and it's a big chunk of change saved on payroll.
Keeping your headcount or hiring humans instead, AI would have to start to cost upwards of $15k/month/developer or more before it costs more than hiring. You're looking at about 4 billion tokens per month before humans start to break even or are cheaper.
True, that was more hypothetical if it got good enough to 2x.
But even taking a more realistic 1.25x (20% time savings) gain, lets say you drop from 500 to 400 devs, you'd have to hit around $4,000/dev/month in token spend before hiring humans again would break even.
Payroll is just expensive, in most companies it's by far the biggest expense. AI still has to cost drastically more before investors would call it out as being worse than increasing headcount, from a pure dollars perspective.
There is no profit, expense, revenue. Those don't matter. Only thing that matters is stock price goes up, and laying off makes stock price go up. When laying off make stock price go down, then laying off stop.
I'm surprised they even had them in a first place. Doesn't Microsoft have a deep partnership with OpenAI? Aren't all Copilot things powered by various GPT models? I would assume the two companies have barter agreements of sorts.
My experience is, Claude Code burns way more tokens compared to other agents, probably to ensure high levels of perceived quality, which is, most of the times not worth the bloat for the user. The bloat works for Anthropic as an advertisement at the cost of your tokens.
What's the point of eating your own dog food when the only thing you are doing is reselling other people's dog food? Microsoft don't have any competing LLM.
> It was part of an effort to get project managers, designers, and other employees to experiment with coding for the first time.
I suspect they weren't as efficient as they could be with token use either. Sounds like they were trying to encourage non-developers to vibe code stuff
I'd argue you have a lot more to worry about with developers as far as token usage goes because they're the ones who know how to rig up these wild workflows where tens of agents simulate an entire software development team. The non-developers are probably going to be sticking more in the realm of iterating via chat.
Lots of these places measure employee token use with managers having dashboards. It seems like performative code production rather than making anything useful.
"incentivize to use as many tokens as possible" = "Upper management knows people dont like change so we are forcing them to come up with ways to use this thing". It does not mean that management will encourage wastefulness in the future, and it also doesnt mean that token usage from now wont be reviewed in the future. Whats to stop them from dinging your performance in november because you wasted a hundred thousand on tokens with nothing to show for it?
Makes sense why Anthropic wants to IPO as soon as possible as the growth right now comes from temporary wastefulness. Makes all the investments more risky.
Microsoft poorly manages token use of most expensive models in a pilot. Then they use that failure to advertise/position their own Github Copilot agents to procurement teams, over the now widely validated Claude Code-based agents.
At least Codex is trying to win validation on merit.
Surely a company as large as Microsoft is actively attempting to build their own models. They couldn't possibly have expected to stake the future of their software development on the conditions of a third party company?
Okay, but what if you're not Microsofts size and don't have and R&D budget large enough to fund development of your own models and tools?
This is a warning to any company, not building their own AI, that AI assisted development could become really expensive really fast and most likely won't pay off. What Microsoft is suggesting is that the current price is to high, but it's still not high enough for e.g. Anthropic to be profitable, or AI coding tools are only as good as the developers using them. So you can't meaningfully do layoffs by replacing the developers with AIs, because the cost is to high.
How does Microsoft plan to fix CoPilot, so that the cost will be so much lower than Claude, that budget overruns won't be a problem for their own customer?
I expect in the next year or so, we'll stop seeing headlines like "Anthropic buys $15b of compute from SpaceX" and we'll start seeing headlines like "Uber's AI department licenses GPT 6.2 as the foundation for their internal model," or something like that.
Smaller companies will have departments that distill larger models into something more specifically manageable and useful for them. At least, that's my personal prediction :)
How would that help with pricing? The cost of hardware is already subsidized to hell and back by investors and that's not dropping costs enough. I'm not concerned about Uber, they are way to big. I'm thinking sub 1000 employees in total and maybe 50 - 100 people in the IT department. Are they just going to be cut off from AI tools, because the cost of running them would ruin the company?
I do think your prediction makes sense, because the AI really isn't the product, it needs to be baked into something and licensing the models saves you the R&D and cost of implementing your own.
The frontier model space costs 1000x as much to develop as the small language models, and is only 1.5 years ahead.
Factually, the frontier models have not paid for themselves. So, if you're MSFT and Apple, you don't need to run in a race where even the winner loses massively.
You can try to train models 1.5 years behind that are highly likely to be profitable, given your market position.
The average person is lagging behind what AI is capable of by 3+ years anyway...
So you can save 1000x on training and 10x on inference and just use SOTA small models.
Why spend $5B training a model that's for sure not going to make $5B (after inference costs) when you can spend $5M building one that WILL make far more than that after inference costs?
At one point there were rumours that they'd do that. They also have the rigts to oAI models for a few more years still, so they could always use that but apparently they're also compute starved (like anyone else).
MSFT does have a frontier AI Lab. My friend works there. I don’t know what they’re doing. But MSFT is one of like 5 entities that actually have the talent and physical infrastructure to compete in model-building.
I think whats funny is that employees were most likely already covering the cost for these tools because they are useful. Companies didn't believe employees were using these tools and now have forced their usage and no longer have the costs subsidized.
Similarly companies seem to reward high token usage as a sign of someone willing to play ball with AI and again have forced higher costs on themselves for people reward hacking or using tokens out of spite.
There is no world where I can put my company’s data through an external site without their express consent and security sign off. I suspect at most companies there’s zero path for people to have been paying for it themselves.
None of the 5 places I have worked is this possible, but they are also all highly regulated industries. Firewalls block virtually everything by default.
Fair, but I assume everything on my work laptop is key logged. Surely they would notice Claude phoning home from my company laptop? I suspect a network rule to look for that traffic is trivial?
The way coding agent work is fantastically wasteful. All the megabytes of code are processed over and over and over, sometimes withing just one session.
There are papers describing KV cache precomputation for commonly used documents (e.g. KVLink), but, of course, it's not a priority for model providers: they'd rather sell you more tokens, also they would rather get to AGI/ASI first than optimize usage of existing models...
I meant caching on a bigger level. If you're an organization with 100 developers each doing 10 sessions a day, you're paying for 10000x tokens in frequently used document even if you had 100% KV cache hits within one session. Apparently that's too costly even for companies with trillion dollar market cap...
Normally KV cache works only if your context prefix is identical, but there are papers which demonstrate documents can be cached between different contexts.
I believe OP is talking about new sessions or after compaction. He’s getting at the fact that LLMs are stateless and have to rediscover your codebase on every new session.
I think so too, otherwise why wouldn't you put that (purported) increased capacity/output into improving your existing products or creating new ones, with the headcount that you already have?
I switched from Anthropic to OpenAI after spending ~$40K in equivalent token costs using Claude over 3 months.
I found Opus 4.7 to be slow and wasteful with token usage. It's shocking how inefficient it is with tasks like bash tool usage and web searching, delegating them to a dozen subagents only to get stuck and never return until you esc and intervene. That, in addition to all of the broken tooling Anthropic built in to limit token usage like the broken monitoring tool made managing Claude a chore. I was happy to pay $200/month for Opus 4.5 when they had more capacity, but 4.7 felt like a huge step back and no longer worth the price and inconvenience.
I remember an OpenAI employee comment on the GPT5.5 release post about how they specifically geared it towards long-horizon tasks and its been a breathe of fresh air in that regard. I have five two-week long sessions going right now and there's been no degradation in performance or efficiency. It's much better at carrying rules/learnings forward even in long-running sessions and grounding/refreshing itself in verified facts when it loses context.
Its funny because in two weeks I've gotten way more done with GPT5.5 with way fewer tokens and way less handholding. I think this goes to show how important tooling and the harness is and how a capable model like Opus 4.7 can be severely handicapped by bad product decisions.
Being able to mange context over long running sessions is a function of the harness, not the model. Are you using Claude Code with GPT5.5? Codex? piclaw? They’ll all have different context management strategies to let you keep going when you would otherwise have filled up context and be forced to stop.
2nd link doesn't work.
That would be a neat tool, to find the original article and see how many levels of AI summary it has gone through, a game of AI telephone!
I had thought about creating something like that for finding comments for articles. For a given article, display links to comments for HN, lobsters, reddit, etc. However, I feel I already waste too much time reading comments. I shouldn't make it easier and more tempting.
Man, maybe it's time for me to give the verge a subscription. There the only ones actually doing any journalism here and a bunch of AI blogs skimming off the top.
> I understand that Microsoft is planning to remove most of its Claude Code licenses and push many of its developers to use Copilot CLI instead. While Claude Code has been a popular addition, it has also undermined Microsoft’s new GitHub Copilot CLI coding tool — a command line version of GitHub Copilot that runs outside of development apps like Visual Studio Code.
And people here are interpreting this as related mainly to the Claude burning too much tokens too quickly and suggesting Microsoft should rather use SomeOtherLLM©?
Is this Hacker News or rather Marketing Wars?
reply