It's a signal that the finance people are becoming more important to the company, but not necessarily a bad thing; it's effectively a more (tax) efficient form of dividends, which isn't very controversial.
Every time I can remember that the finance people have become more important to a company, it has led to the disappearance of the internal culture geared towards excellence that got a company to that point in the first place.
I guess it's simply the externalities issue. Finance people optimize finance, which among other things results in improved efficiency. But despite best intentions to map out all incentives that matter, it always fails to consider some aspects. Focusing on short-term profits, ignoring privacy, security or pollution because it's free, lobbying for favourable legislation that hampers competitors, etc.
These are things that don't show in a spreadsheet unless you're explicitly incentivized to look at them. But that's never the case because the number of KPIs is always finite while there are infinitely many aspects that could potentially be subverted.
It’s largely true. I believe Steve Jobs had some talk about this phenomenon; basically, at some point, the scale of an organization means the product people make less of an impact than the sales / finance people, and they slowly take over.
Then over the span of a few decades, what’s left is a shallow organization without real innovation.
That's what they're afraid of, and that's one of the reasons why they're doing the big management reorganization - too many managers leads to lots of overhead instead of excellence.
But both dividends and stock buybacks are terrible and really shouldn't exist. In a proper market, competition is so fierce that you cannot afford dividends / stock buybacks because your competitors will put all their money towards R&D and retaining & attracting the best personnel.
Then again, this has been going on for decades. Businesses used to be about being the best for your customers and personnel. But it's all become about sticking it to everyone for the benefit of the shareholders.
It’s even less of a feature, Claude Code already has subagents; this new feature just ensures Claude Code actually uses this for implementation.
imho the plans of Claude Code are not detailed enough to pull this off; they’re trying to do it to preserve context, but the level of detail in the plans is not nearly enough for it to be reliable.
I agree with this. Any time I make a plan I have to go back and fill it in, fill it in, what did we miss, tdd, yada yada. And yes, I have all this stuff in CLAUDE.md.
You start to get a sense for what size plan (in kilobytes) corresponds to what level of effort. Verification adds effort, and there's a sort of ... Rocket equation? in that the more infrastructure you put in to handle like ... the logistics of the plan, the less you have for actual plan content, which puts a cap on the size of an actionable plan. If you can hit the sweet spot though... GTFO.
I also like to iterate several times in plan mode with Claude before just handing the whole plan to Codex to melt with a superlaser. Claude is a lot more ... fun/personable to work with? Codex is a force of nature.
Another important thing I will do is now that launching plans clear context, it's good to get out of planning mode early, hit an underspecified bit, go back into planning mode and say something like "As you can see the plan was underspecified, what will the next agent actually need to succeed?" and iterate that way before we actually start making moves. This is made possible by lots of explicit instructions in CLAUDE.md for Claude to tell me what it's planning/thinking before it acts. Suppressing the toolcall reflex and getting actual thought out helps so much.
It’s moving fast. Just today I noticed Claude Code now ends plans with a reference to the entire prior conversation (as a .jsonl file on disk) with instructions to check that for more details.
Not sure how well it’s working though (my agents haven’t used it yet)
Interesting about the level of detail. I’ve noticed that myself but I haven’t done much to address it yet.
I can imagine some ideas (ask it for more detail, ask it to make a smaller plan and add detail to that) but I’m curious if you have any experience improving those plans.
Effectively it tries to resolve all ambiguities by making all decisions explicit — if the source cannot be resolved to documentation or anything, it’s asked to the user.
It also tries to capture all “invisible knowledge” by documenting everything, so that all these decisions and business context are captured in the codebase again.
Which - in theory - should make long term coding using LLMs more sane.
The downside is that it takes 30min - 60min to write a plan, but it’s much less likely to make silly choices.
Have you tried the compound engineering plugin? [^1]
My workflow with it is usually brainstorm -> lfg (planning) -> clear context -> lfg (giving it the produced plan to work on) -> compound if it didn’t on its own.
> The downside is that it takes 30min - 60min to write a plan
Oof you weren't kidding. I've got your skills running on a particularly difficult problem and it's been running for over three hours (I keep telling it to increase the number of reviews until its satisfied).
Yeah I’m working on some improvements in this area, should make things faster. But yeah I’ve frequently had 1h-2h planning sessions as well, depending upon the complexity of the task.
I have had good success with the plans generated by https://github.com/obra/superpowers I also really like the Socratic method it uses to create the plans.
I iterate around issues. I have a skill to launch a new tmux window for worktree with Claude in one pane and Codex in another pane with instructions on which issue to work on, Claude has instructions to create a plan, while Codex has instructions to understand the background information necessary for this issue to be worked on. By the time they're both done, then I can feed Claude's plan into Codex, and Codex is ready to analyze it. And then Codex feeds the plan back to Claude, and they kind of ping pong like that a couple times. And after a certain or several iterations, there's enough refinement that things usually work.
Then Claude clears context and executes the plan. Then Codex reviews the commit and it still has all the original context so it knows what we have been planning and what the research was about the infrastructure. And it does a really good job reviewing. And again, then they ping pong back and forth a couple times, and the end product is pretty decent.
Codex's strength is that it really goes in-depth. I usually do this at a high reasoning effort. But Codex has zero EQ or communication skills, so it works really well as a pedantic reviewer. Claude is much more pleasant to interact with. There's just no comparison. That's why I like planning with Claude much more because we can iterate..
I am just a hobbyist though. I do this to run my Ansible/Terraform infrastructure for a good size homelab with 10 hosts. So we actually touch real hardware a lot and there's always some gotchas to deal with. But together, this is a pretty fun way to work. I like automating stuff, so it really scratches that itch.
That’s what I also thought when I saw this incident. I wonder if there’s something up internally at Cloudflare or that it was always like this.
I feel like something such as a route leak should not be something that happens to Cloudflare. I’m surprised they set their systems up to allow this human error.
John left in April last year I think so it probably isn't directly related, so please take my comment in jest, but still it is worrisome, CF is in many ways 'too big to fail' and if this really becomes a regular thing it is going to cause a lot of people focused on their 'nines' to be pissed off.
One thing to their credit though: BGP is full of complexity and it definitely isn't the first time that something like this goes wrong, it is just that at CF scale the impact is massive so there is no room for fuckups. But doing this sort of thing right 100% of the time is a really hard problem, and I'm happy I'm not in any way responsible for systems this important.
Whoever is responsible learned a lot of valuable lessons today (you hope).
The older I get, the less I buy into "too big to fail" arguments. I now view it as "can't fail soon enough". The sooner it breaks down, the sooner something better will supplant it.
This last sentiment holds true generally since organizations no longer subject to meaningful competition inevitably squat on their laurels and stop excelling at the things they used to be good at. We've seen it everywhere - Boeing, Google, Microsoft (with OS's), etc.
There was never much of an argument behind "too big to fail", it is generally a euphemism for upper-class welfare. In a more realist world, "too big to fail" is a mis-statement of "too risky to keep". Everything fails eventually and keeping incentives aligned relies on having a mechanism - failure - to flush out incompetents.
> The sooner it breaks down, the sooner something better will supplant it.
That's not always possible, because the counterparty - aka threat actors - is always growing bigger, and you practically need to be the size of Cloudflare, Akamai or the Big 3 cloud providers to be able to weather attacks. You need to have big enough pipes to data centers and exchange points worldwide, otherwise any sufficiently motivated attacker can just go and swamp them, but big pipes are helluvalot expensive so you need to have enough large and financially capable customers.
That's also why Cloudflare has expanded their offerings so much (e.g. Zero Trust), they need to have their infrastructure at some base load to economically justify it.
And that's also why Cloudflare will not be kicked off the throne any time soon. First of all, the initial costs to set up a competitor are absurdly high, second, how is a competitor supposed to lure large long term customers away from CF?
Any case, the real "fix" to Cloudflare being too-big-to-fail isn't building up competitors, it's getting the bad actors off of the Internet. Obviously that means holding both enemy (NK, Russia, China) and frenemy (India, Turkey) nations accountable, but it also means cleaning up shop at home - the aforementioned nation states and their botnet operators rely on an armada of hacked servers, ordinary computers and IoT devices in Western countries to carry out the actual work. And we clearly don't do anywhere near enough to get rid of these. I 'member a time when writing an abuse@ mail report that this would be taken seriously and the offender being disconnected by their ISP. These days, no one gives a fuck.
"Threat actor" is a relative definition, because for Italy the Cloudflare CEO was a "threat actor" who openly threatened availability of their systems.
Cloudflare knows they are just a glorified firewall + CDN that's why they desperately push into edge computing and getting these dozens of features.
I believe the constitution is part of its training data, and as such its impact should be consistent across different applications (eg Claude Code vs Claude Desktop).
I, too, notice a lot of differences in style between these two applications, so it may very well be due to the system prompt.
Exactly, the current spend on LLMs is based on extremely high expectations and the vendors operating at a loss. It’s very reasonable to assume that those expectations will not be met, and spending will slow down as well.
Nvidia’s valuation is based on the current trend continuing and even increasing, which I consider unlikely in the long term.
> Nvidia’s valuation is based on the current trend continuing
People said this back when Folding@Home was dominated by Team Green years ago. Then again when GPUs sold out for the cryptocurrency boom, and now again that Nvidia is addressing the LLM demand.
Nvidia's valuation is backstopped by the fact that Russia, Ukraine, China and the United States are all tripping over themselves for the chance to deploy it operationally. If the world goes to war (which is an unfortunate likelihood) then Nvidia will be the only trillion-dollar defense empire since the DoD's Last Supper.
China is restricting purchases of H200s. The strong likelihood is that they're doing this to promote their own domestic competitors. It may take a few years for those chips to catch up and enter full production, but it's hard to envision any "trillion dollar" Nvidia defense empire once that happens.
What’s the definition of linkbait in this case? I can’t find a good definition online, as the best I can find is “content designed to attract backlinks”, which this does not appear to be / isn’t related to the title?
My guess as to why it was changed was that it doesn't really work as a pithy quip the moment you actually think about it.
OpenAI is introducing ads on their purported path from "AI" to "AGI"... hence the "A" was already accounted for in the acronym. If only "Ads" started with "G"!
reply