Having listened to him give a couple of talks I am always struck by how much he talks and writes like Claude code.
I don’t think he farms out all of his writing and talking to LLMs. I don’t think Claude code was trained to emulate him or anything.
I think his “voice” has been filed LLM smooth by years of agent based interactions. He claims Anthropic engineers use an average of 500+ agents a day. They are human interfaces to token generators more than human to human communication. They are picking up the tendencies of their most frequent communication partner.
Unfortunately, I see Claude being wrong often enough that I view it as a faulty narrator. Often helpful, sometimes totally full of it.
And now I subconsciously apply this filter to anything that sounds like Claude.
Makes me nervous that my voice may be becoming that of a faulty narrators.
Higher level thinking has fewer constraints on being wrong than lower level subconscious thinking. Low level stuff tends to have some repetitive and evolutionary basis. Higher level thought tends to be more one off, less evidence based, and more exploratory until we get to the point of being an expert around particular concepts.
A huge amount of human thinking is being a parrot and repeating statements without a deeper understanding. For example when I say 1+1=2 I'm not thinking about it. If I say 638483+949948= calculation is necessary.
So in this, your voice has always been one of a somewhat faulty narrator, LLMs just may be further degrading the quality of the narration.
Lets say you have a colleague named Bob. Bob is a pretty good engineer, but he is as confident when he doesn’t know as when he knows.
You can never tell his degree of confidence in his answer. For low-level things he is pretty good but for critical high-level things it is more ambiguous.
You work around this by checking Bob’s work. Having formal verification steps to check what he says.
Now let’s say you meet Bob’s twin brother Bill. Who talks and sounds just like Bob. He makes fairly grand high-level statements that you can neither test nor verify.
>"They are human interfaces to token generators more than human to human communication. They are picking up the tendencies of their most frequent communication partner."
I see this happening at work, too. All communication and collaboration between engineers has broken down. Juniors don't have questions anymore. Seniors don't discuss architecture anymore. Everyone is just off in their own world of feature work with the only interaction coming at merge time. Really depressing what has happened to our trade.
That isn't humanly possible, you will blow out your neurotransmitters and not be able to think. Only being semi serious, but we can't make that many decisions a day.
What does that mean though? Does it mean there are more than 500 different "agent" systems within Anthropic and employees interact with most of those in a daily basis?
> Something that people learn quickly when they work with me is that my approach to pretty much every problem is:
> ...
> 6. Act with urgency to achieve the goal
If everything is urgent, then nothing is. This guy sounds miserable to work for and with. Assuming this is accurate and not just hyperbole, he is essentially saying he has no prioritization skills because everything is urgent. I think most people who have been around the block have worked with people like this, and unbeknownst to them, their coworkers develop a default snooze button associated with most of their requests and projects.
If someone literally publishes on the internet that they require their employees to act like every problem is urgent, it's not clear why you think we shouldn't take them at their word. It's a pretty common thing for bad managers to expect people to work full-tilt 100% of the time rather than understanding that tends to burn poorly out. I'm honestly not sure how someone could spend a decent amount of time in this industry and not be able to recognize that pattern and have thoughts about how to avoid it unless they just don't really care about it much, so while you're entitled to disagree, it's hard not to get the impression that being so affronted by the parent comment pointing out that the blog post doesn't seem to address the very real pattern of phrases like the article uses being euphemisms for pretty horrible work cultures that you just also don't particularly care about it.
Personally, I've seen far too many talented people in tech be worked too hard for too long by management that either is too clueless or too uncaring to the point where it can take years for their health and happiness to recover not to find it extremely alarming when someone says stuff like "everything is urgent". The likelihood that they mean pretty much exactly what they say rather than mean something more nuanced is high enough that the fact that don't seem concerned about the implication says more than enough.
> Heavens. Is it possible you're reading into it a bit?
Perhaps, which is why I hedged with the possibility that it is mere hyperbole. However, I think what did it for me was his statement that he provides and expects feedback to anyone not following this recipe, which is at utterly without nuance. Incidentally, you can execute with focus and intention without urgency and arguably be more effective over the long haul.
This list also makes zero sense. It says to gather missing information after gathering all known information but before defining a problem. How could you even know what information is missing, much less information that is important, without knowing the problem? And if you're to gather missing information, that you somehow know about, doesn't that make it part of gathering known information? The rest of the list is just as dumb. This list is like a middle schooler was asked to come up with a problem solving framework.
This guy is usually all over threads in which he gets to show off his internal knowledge of Claude Code. I'm sure he'll be here after getting clowned.
The entire Bay Area seems like the most insufferable people known to man. These people have zero introspection.
> It says to gather missing information after gathering all known information but before defining a problem. How could you even know what information is missing, much less information that is important, without knowing the problem?
The post’s description of these steps is reasonable IMO. I’d write it as “put in order the relevant information you have” and “find the information which you know you need but don’t have at hand”.
By way of bad analogy, one could imagine writing up a document first off the top of your head, then filling in more of the document based on the documentation of the relevant systems, corresponding to these two steps the author describes.
I can't say I agree with forcing your team to use your own personal "framework" of approaching a problem or else you'll get "feedback".
> Sometimes I will give feedback to people when they are missing steps in the framework, or are poorly executing some of the steps. I expect the same feedback in return.
I also don't like that urgency is built in as the standard process either, no wonder everyone is burnt out.
Acting with urgency is a bit at odds with discovering flaws in your plan. If you're sprinting you're less likely to notice smells and things that are inelegant, more likely to paper them over. That said, there is a time for urgency. Just not every single task.
This seems strongly aligned with the Amazon doc-writing and decision making process. I found it to be unusually effective as a business process, and took it with me when I left to my current role.
Making thoroughly informed decisions and iterating on a decision doc before committing to a direction and plan is better than every alternative I’ve ever observed in my career.
The criticism I’ve read thus far on this thread seems unwarranted. I give the same kind of feedback to my mentees when their work product or process could use improvement.
I worked for AWS for several years, eventually leaving during the big "return" to office push because my geographically distributed team working on a product that didn't exist back before the remote period was told we had a month to either move to one of three cities where our team was allowed to work out of our transfer to a team in our area. My wife had an autoimmune condition that would put her at risk if I commuted, so I asked my manager what processes there were for exemptions, but he literally hadn't been told anything by the higher-ups and that as far as he was aware, there were no processes defined at all and we'd have to just try to talk with HR and upper management to try to figure something out. I didn't think that the company expecting me to rush to figure it out when they were the ones who put an artificially constrained timeline on us to either abandon what we had been working on it uproot our lives, so I ended to just giving my notice a couple of days later instead.
My point here is that companies that love to have extremely formal processes around for to handle technical things are not necessarily less likely to have extremely arbitrary decisions without any process for recourse defined when it comes to how employees get treated. If someone describes a technical process that they say they use for everything that a literal reading seems pretty dubious in regards to things like burnout or micromanagement, I don't think it's that crazy to recognize that the reality is probably at least as bad as the obvious implication. Sure, they're not directly saying "I overwork my employees by imposing short deadlines on everything and I nitpick their processes if they different from my own", but there are enough managers who do act this way that it's kind of hard to think someone who cared about being perceived as saying that wouldn't go out of their way to clarify where the nuance is if it truly exists.
> Making thoroughly informed decisions and iterating on a decision doc before committing to a direction and plan is better than every alternative I’ve ever observed in my career.
My last company did this. It usually devolved into a design-by-committee full of compromises to make various stakeholders happy and often yielded a worse artifact.
It became more effective once people got burnt out on the process and most stakeholders stopped caring and started rubber stamping, allowing the one or two people willing to put in the energy to come up with something coherent.
I still am mind blown at how bad CC is as software. It’s just not that hard of a problem. I get that harnesses aren’t trivial, but they’re not insane either. And the fact that it’s running on a JS runtime (that they bought!) is also crazy. Why not Go/BubbleTea? Why not literally anything native? It makes no sense
I don’t want to be a jackass, but it’s hard to take anything Boris says seriously when he’s headed up such weird software. And it’s not like resourcing or money is an issue for them. If Claude was so damn good, why does CC suck?
I dont't really know stats about such things, but I assume the adoption of CC has been insane compared to almost anything, and is also like two years old? Yeah, not surprised.
But also, if they all use 500 agents daily and latest models can "one shot" everything, one would think 5k issues are fixed in no time...
Havent used CC in some time, worked fine last time I tried.
> But also, if they all use 500 agents daily and latest models can "one shot" everything, one would think 5k issues are fixed in no time...
IMO, CC should be the poster child of what LLM coding should "feel" like. If things are so good why are there so many issues? Why can't they get a handle like other well-run projects? As you said, they have unlimited tokens so this project should be close to pristine as much as possible.
Those numbers don't really mean anything. Claude Code has millions of users, and that issue forum is the most obvious place for them to ask questions or request features.
If there were 11,000 open and confirmed bugs then yeah, that would mean the software is bad.
No one knows how many bugs has because Anthropic auto-closes Issues (or at least used to) if there isn't someone constantly pinging the issue every two weeks. I've contributed to several issues that were confirmed by several other people, which were they were auto-closed after people gave up confirming the problem without any response from Anthropic. They seemingly don't care if it isn't on fire or at least smoldering heavily, which seems really bad in my book if you're trying to make even reasonably good software.
I'm surprised you think Claude Code is good software. I find it so hard to use because it is fundamentally constrained as a TUI. It is buggy, clunky and slow. It is sooo slow.
One example of what is particularly bad with it: tool calls and progress. Codex UI does it so much better. The way in which CC waits for a task to complete etc is really poorly done compared to Codex.
Another thing that's missing: no way to continue a side chat in Claude Code - this is easy in Codex with /side and it is super usefu.
Claude Code has /btw which I think is the equivalent of /side in Codex.
I don't think Claude Code is perfect software, but I don't think fact that Codex has a slightly nicer implementation of certain patterns makes Claude Code bad software.
I’ve rolled my own harness in Go. I have a rule to not let the production LoC exceed 50k lines. It is _very_ nice for my use cases. DeepSeek Flash V4.1 often performs at the level of GPT 5.6 Sol for non-orchestration tasks (programming and maths).
I add features and tweak it all the time: You just need to get comfortable with spending 3 hours in a chat session every 3 weeks to think hard about how to simplify whatever slop you didn’t simplify the last time you did this :)
Steps 1–5 of his framework are increasingly formal ways of saying “figure out what’s going on before doing something,” followed by step 6: “then do it fast”
If you do this step well, in my experience, the next steps fall into place quickly.
My version: accurately naming the problem is the essence of problem solving. There are things you usually need to do before you can accurately name the problem, because most problems aren’t presented to you in textbook form.
Once properly defined and framed, it’s obvious how to solve most problems, most of the time.
If you’ve ever worked with someone amazingly good at accurately naming the problem, it’s hard to unsee it. This is one of the hidden talents of the most effective people I’ve met.
It is strange how people who say they are "often wrong", never quite sound like they take that into consideration when delivering opinions or thought pieces - I see it fairly often in tech. Anyways; nothing against Boris, but Anthropic/Claude has certainly been the first company (& product) too annoying for me to use.
It seems strange to write a blog post about being proven wrong, without a single example of being wrong. Especially with all the current Claude.md drama.
Everyone loves being wrong in the abstract - this reads more like a self-congratulation than a genuine reflection.
1) I start to believe my own hype because I am so consistently correct- which blinds me to other ideas if I don't actively take steps to humble myself.
(after all, I won't be right very often if I stop listening, as that's part of why I am right, because I listen to people and have a huge amount of context)
and
2) I am able to quickly come to a conclusion which makes people uneasy, as they think I haven't considered all the points even though I have- and it's also true that people have a bias to inaction unless it's an urgent issue.. this is how things get stuck in committees and meetings and continually get kicked down the road because it's difficult to get everyone together and there's a million reasons to keep deferring meetings.
What I'm trying to say is, don't sit there thinking that "if only I could make correct decisions"- because even if you are quick, correct and decisive: you will ruffle feathers.
(at the risk of sounding arrogant: i'm not saying I'm right about everything, like the red baron, I pick my battles very carefully).
This guy doesn't think deeply about any problem, possibly because he has never had to and probably because he has never wanted to. He would probably make a good residential plumber.
You must not be a home owner because good residential plumbers are definitely out there solving the “hard problems” that techies love to bloviate about every day.
And yet there's none of this uncertainty or caution with his posts that then go on to have massive ripple effects because of his position at Anthropic and the marketting related to claude.
I personally have a lot of anger and frustration with many people in the ai hypesphere that are just mindlessly frolicking around without a care in the world, happy to speak into the megaphone offered by masses that are in a rat-race to avoid some AI dystopian hellscape that keep getting painted by these thought leaders... and then going "oopsies... I am just human guys... y so mad?!"
I think someone can follow "good practices" with a team and get to a bad place, especially with such a novel product (as AI coding agents).
But taking Claude Code as the product of this style of thinking -- who is Claude Code for? Is it for everyone in the world? Well, if you look at the feature velocity, it seems like the answer is intended to be yes ... Claude Code is trying to solve every problem in software development in the world, all at the same time.
So I question the "user model" here.
Here's another thing that is true about Claude Code: it's among the most inconsistent and buggy pieces of software I've ever encountered.
- You can move the cursor with the mouse in the composer, but not in AskUserQuestion?
- When agents spawn subagents, the model name is inherited from the main agent, and seemingly none of the (4! yes, 4!) subagent tools seem to get this right (except for Explore, which seems to be fixed to a weaker model)
- Sometimes, when my usage limit halts, my agents will pick up when it refreshes (within ~2 hours or something) ... other times, nope -- even within the usage limit?
This is a sampling of my own experiences using this thing frequently. Are these sorts of details not important? Maybe not: I'm not at the level of this team, and may never be.
But I think it's a reflection of agentic engineering ... a somewhat embarrassing one, from my perspective. It paints a picture of a team who can't quite get the details right, even with the assistance of purported extremely powerful AI tools, even internal ones which we don't have access to?
I think when people look back on 2025 -- Boris is going to have his name right there in the books ... Claude Code, coding agents -- Anthropic (& Boris + team) made the first move.
But now it's 2026, and people know how harnesses work, and heavy lies the crown.
Heaps of comments here but nobody has pointed out he defined the goal after defining the problem. That's cheating.
Step one is to define the goal. Then you work out how to reach the goal. (Gather information and form a hypothesis) Then you take steps toward the goal. (Test your hypothesis to make sure you stepped in the right direction.)
The most annoying thing about this is that the steps are out of order unless the definition of problems and goals are intermixed
1. define a goal
2. understand information and gaps in information
3. define and prioritize the problems to achieve that goal
4. identify potential solutions for top problems and prioritize
5. measure success
and do all of it with urgency (magically managers want everything done with urgency)
This is what happens when you let things get to your head. You work on a (highly inefficient, often broken) piece of extremely basic software by prompting a slot machine and hoping it works. Calm down and stop forcing your shitty framework on employees, you're the most replaceable cog of all.
Yeah? A very basic process, most people use some variation of this implicitly without talking about it.
This sounds like a slightly narcissistic manager who thinks people are doing it wrong if they don't act like small copies of him. There are multiple ways to reach the same outcome and a lot depends on your information. E.g. I typically have a mental model of a system in my head, meaning when some problem needs fixing very likely I already know where it would need fixing and already think about the various future implications arising from a different fix. A point that is totally absent from that framework.
Being a senior dev myself I have seen enough good software turn bad to know that seemingly innocent technological decisions can come with huge and lasting implications. The fact that this is missing here is speaking volumes about the lack of experience at display here.
Your task as a manager isn't to create small copies of yourself. Your task is to know each persons weaknesses and strengths and compose the work in such chunks that the weaknesses have little effect, while the strengths multiply. For this you will first of all have to trust your people and lead them to discover certain ideas themselves.
E.g. if you feel someone always jumps gun-ho I to the task without doing the research, just tell them to give you the research first. Do that a few times and they might realize how useful that is.
Having listened to him give a couple of talks I am always struck by how much he talks and writes like Claude code.
I don’t think he farms out all of his writing and talking to LLMs. I don’t think Claude code was trained to emulate him or anything.
I think his “voice” has been filed LLM smooth by years of agent based interactions. He claims Anthropic engineers use an average of 500+ agents a day. They are human interfaces to token generators more than human to human communication. They are picking up the tendencies of their most frequent communication partner.
Unfortunately, I see Claude being wrong often enough that I view it as a faulty narrator. Often helpful, sometimes totally full of it.
And now I subconsciously apply this filter to anything that sounds like Claude.
Makes me nervous that my voice may be becoming that of a faulty narrators.
To err is human.
Higher level thinking has fewer constraints on being wrong than lower level subconscious thinking. Low level stuff tends to have some repetitive and evolutionary basis. Higher level thought tends to be more one off, less evidence based, and more exploratory until we get to the point of being an expert around particular concepts.
A huge amount of human thinking is being a parrot and repeating statements without a deeper understanding. For example when I say 1+1=2 I'm not thinking about it. If I say 638483+949948= calculation is necessary.
So in this, your voice has always been one of a somewhat faulty narrator, LLMs just may be further degrading the quality of the narration.
Lets say you have a colleague named Bob. Bob is a pretty good engineer, but he is as confident when he doesn’t know as when he knows.
You can never tell his degree of confidence in his answer. For low-level things he is pretty good but for critical high-level things it is more ambiguous.
You work around this by checking Bob’s work. Having formal verification steps to check what he says.
Now let’s say you meet Bob’s twin brother Bill. Who talks and sounds just like Bob. He makes fairly grand high-level statements that you can neither test nor verify.
How much do you intuitively distrust him?
Left, right, higher, lower, male, female. People love to explain the brain by dividing it in half, or into poles.
Nothing more human than to label, simplify and categorize, even in the absence of higher quality information.
>"They are human interfaces to token generators more than human to human communication. They are picking up the tendencies of their most frequent communication partner."
I see this happening at work, too. All communication and collaboration between engineers has broken down. Juniors don't have questions anymore. Seniors don't discuss architecture anymore. Everyone is just off in their own world of feature work with the only interaction coming at merge time. Really depressing what has happened to our trade.
> He claims Anthropic engineers use an average of 500+ agents a day.
What are they doing all day with that, I wonder?
Well, trying to make RSI for one so humans don't have to do that work.
LLMs are still a very new technology in relation to human timeliness. There is a whole lot of exploration to be done.
I mean, in relation to human timelines, so are computers in general. It's not close why you think that this is unique to LLMs.
Does that mean 500 separate agents or 500 agent sessions I wonder.
That isn't humanly possible, you will blow out your neurotransmitters and not be able to think. Only being semi serious, but we can't make that many decisions a day.
the former
What does that mean though? Does it mean there are more than 500 different "agent" systems within Anthropic and employees interact with most of those in a daily basis?
> Something that people learn quickly when they work with me is that my approach to pretty much every problem is: > ... > 6. Act with urgency to achieve the goal
If everything is urgent, then nothing is. This guy sounds miserable to work for and with. Assuming this is accurate and not just hyperbole, he is essentially saying he has no prioritization skills because everything is urgent. I think most people who have been around the block have worked with people like this, and unbeknownst to them, their coworkers develop a default snooze button associated with most of their requests and projects.
Heavens. Is it possible you're reading into it a bit? He's miserable to work with? From this one blog post?
A more generous read, or, at least the reading I took: "once we (think we) know what we're doing, we violently execute."
So, "boo" on you. This guy sounds awesome.
If someone literally publishes on the internet that they require their employees to act like every problem is urgent, it's not clear why you think we shouldn't take them at their word. It's a pretty common thing for bad managers to expect people to work full-tilt 100% of the time rather than understanding that tends to burn poorly out. I'm honestly not sure how someone could spend a decent amount of time in this industry and not be able to recognize that pattern and have thoughts about how to avoid it unless they just don't really care about it much, so while you're entitled to disagree, it's hard not to get the impression that being so affronted by the parent comment pointing out that the blog post doesn't seem to address the very real pattern of phrases like the article uses being euphemisms for pretty horrible work cultures that you just also don't particularly care about it.
Personally, I've seen far too many talented people in tech be worked too hard for too long by management that either is too clueless or too uncaring to the point where it can take years for their health and happiness to recover not to find it extremely alarming when someone says stuff like "everything is urgent". The likelihood that they mean pretty much exactly what they say rather than mean something more nuanced is high enough that the fact that don't seem concerned about the implication says more than enough.
I was literally writing almost this exact post but decided to reload the page first.
Developers get triggered about managers breathing down their neck when they read things like this that’s why you get such strong reactions.
How often do people that make this kind of assertion end up themselves being the ones that are miserable to be around I wonder
> Heavens. Is it possible you're reading into it a bit?
Perhaps, which is why I hedged with the possibility that it is mere hyperbole. However, I think what did it for me was his statement that he provides and expects feedback to anyone not following this recipe, which is at utterly without nuance. Incidentally, you can execute with focus and intention without urgency and arguably be more effective over the long haul.
I disagree, on the principle of "slow is smooth, smooth is fast"
and similarly, the need for slack in software teams
This principle is basically a movie quote, though.
No it's not, though?
This list also makes zero sense. It says to gather missing information after gathering all known information but before defining a problem. How could you even know what information is missing, much less information that is important, without knowing the problem? And if you're to gather missing information, that you somehow know about, doesn't that make it part of gathering known information? The rest of the list is just as dumb. This list is like a middle schooler was asked to come up with a problem solving framework.
This guy is usually all over threads in which he gets to show off his internal knowledge of Claude Code. I'm sure he'll be here after getting clowned.
The entire Bay Area seems like the most insufferable people known to man. These people have zero introspection.
> It says to gather missing information after gathering all known information but before defining a problem. How could you even know what information is missing, much less information that is important, without knowing the problem?
The post’s description of these steps is reasonable IMO. I’d write it as “put in order the relevant information you have” and “find the information which you know you need but don’t have at hand”.
By way of bad analogy, one could imagine writing up a document first off the top of your head, then filling in more of the document based on the documentation of the relevant systems, corresponding to these two steps the author describes.
I can't say I agree with forcing your team to use your own personal "framework" of approaching a problem or else you'll get "feedback".
> Sometimes I will give feedback to people when they are missing steps in the framework, or are poorly executing some of the steps. I expect the same feedback in return.
I also don't like that urgency is built in as the standard process either, no wonder everyone is burnt out.
> 6. Act with urgency to achieve the goal
Acting with urgency is a bit at odds with discovering flaws in your plan. If you're sprinting you're less likely to notice smells and things that are inelegant, more likely to paper them over. That said, there is a time for urgency. Just not every single task.
But if you don't only surround yourself with people who think exactly like you, how will you ever recruit an army of yes men and women?
I'll rent an army of yes agents instead.
It’s basically OODA loop, which is itself just as descriptive of how people act and adjust than it is prescriptive.
How else would you describe iteratively solving problems?
Interestingly enough the problem with most of problem solving is identifying the right problem to solve.
It may sound weird but OODA starts making sense once you’ve identified the right problem to focus on.
In its original definition it was based on dogfights. There, your problem is clearly defined.
and the fun question of how do you find good problems is very hard to systematize!
These steps, but in many other arbitrary orders, some with feedback to previous steps
Nothing more annoying than a manager with a personal know-it-all framework requiring people to follow it or else they get "feedback".
He's often wrong, don't you know.
Just not about your performance review.
This seems strongly aligned with the Amazon doc-writing and decision making process. I found it to be unusually effective as a business process, and took it with me when I left to my current role.
Making thoroughly informed decisions and iterating on a decision doc before committing to a direction and plan is better than every alternative I’ve ever observed in my career.
The criticism I’ve read thus far on this thread seems unwarranted. I give the same kind of feedback to my mentees when their work product or process could use improvement.
I worked for AWS for several years, eventually leaving during the big "return" to office push because my geographically distributed team working on a product that didn't exist back before the remote period was told we had a month to either move to one of three cities where our team was allowed to work out of our transfer to a team in our area. My wife had an autoimmune condition that would put her at risk if I commuted, so I asked my manager what processes there were for exemptions, but he literally hadn't been told anything by the higher-ups and that as far as he was aware, there were no processes defined at all and we'd have to just try to talk with HR and upper management to try to figure something out. I didn't think that the company expecting me to rush to figure it out when they were the ones who put an artificially constrained timeline on us to either abandon what we had been working on it uproot our lives, so I ended to just giving my notice a couple of days later instead.
My point here is that companies that love to have extremely formal processes around for to handle technical things are not necessarily less likely to have extremely arbitrary decisions without any process for recourse defined when it comes to how employees get treated. If someone describes a technical process that they say they use for everything that a literal reading seems pretty dubious in regards to things like burnout or micromanagement, I don't think it's that crazy to recognize that the reality is probably at least as bad as the obvious implication. Sure, they're not directly saying "I overwork my employees by imposing short deadlines on everything and I nitpick their processes if they different from my own", but there are enough managers who do act this way that it's kind of hard to think someone who cared about being perceived as saying that wouldn't go out of their way to clarify where the nuance is if it truly exists.
> Making thoroughly informed decisions and iterating on a decision doc before committing to a direction and plan is better than every alternative I’ve ever observed in my career.
My last company did this. It usually devolved into a design-by-committee full of compromises to make various stakeholders happy and often yielded a worse artifact.
It became more effective once people got burnt out on the process and most stakeholders stopped caring and started rubber stamping, allowing the one or two people willing to put in the energy to come up with something coherent.
I still am mind blown at how bad CC is as software. It’s just not that hard of a problem. I get that harnesses aren’t trivial, but they’re not insane either. And the fact that it’s running on a JS runtime (that they bought!) is also crazy. Why not Go/BubbleTea? Why not literally anything native? It makes no sense
I don’t want to be a jackass, but it’s hard to take anything Boris says seriously when he’s headed up such weird software. And it’s not like resourcing or money is an issue for them. If Claude was so damn good, why does CC suck?
Claude Code is also offered as an SDK, you can build custom (customized) harness on top of what essentially is Claude Code. https://code.claude.com/docs/en/agent-sdk/overview
They're a small bootstrapped startup, give them some grace.
I'll be honest, I don't actually understand what people mean when they say Claude Code is bad software.
Seems pretty good to me. Presumably this is about the TUI version?
Do you really think that having this many issues is justifiable/ok?
https://github.com/anthropics/claude-code/issues
I dont't really know stats about such things, but I assume the adoption of CC has been insane compared to almost anything, and is also like two years old? Yeah, not surprised.
But also, if they all use 500 agents daily and latest models can "one shot" everything, one would think 5k issues are fixed in no time...
Havent used CC in some time, worked fine last time I tried.
> But also, if they all use 500 agents daily and latest models can "one shot" everything, one would think 5k issues are fixed in no time...
IMO, CC should be the poster child of what LLM coding should "feel" like. If things are so good why are there so many issues? Why can't they get a handle like other well-run projects? As you said, they have unlimited tokens so this project should be close to pristine as much as possible.
Those numbers don't really mean anything. Claude Code has millions of users, and that issue forum is the most obvious place for them to ask questions or request features.
If there were 11,000 open and confirmed bugs then yeah, that would mean the software is bad.
No one knows how many bugs has because Anthropic auto-closes Issues (or at least used to) if there isn't someone constantly pinging the issue every two weeks. I've contributed to several issues that were confirmed by several other people, which were they were auto-closed after people gave up confirming the problem without any response from Anthropic. They seemingly don't care if it isn't on fire or at least smoldering heavily, which seems really bad in my book if you're trying to make even reasonably good software.
> Those numbers don't really mean anything
IMO this is very dismissive. One example of probably many more, where software is used by millions and yet doesn't have this much being reported.
https://github.com/curl/curl/issues
curl isn't end-user software, and has a very small, well defined surface area.
I'm surprised you think Claude Code is good software. I find it so hard to use because it is fundamentally constrained as a TUI. It is buggy, clunky and slow. It is sooo slow.
One example of what is particularly bad with it: tool calls and progress. Codex UI does it so much better. The way in which CC waits for a task to complete etc is really poorly done compared to Codex.
Another thing that's missing: no way to continue a side chat in Claude Code - this is easy in Codex with /side and it is super usefu.
Claude Code has /btw which I think is the equivalent of /side in Codex.
I don't think Claude Code is perfect software, but I don't think fact that Codex has a slightly nicer implementation of certain patterns makes Claude Code bad software.
I’ve rolled my own harness in Go. I have a rule to not let the production LoC exceed 50k lines. It is _very_ nice for my use cases. DeepSeek Flash V4.1 often performs at the level of GPT 5.6 Sol for non-orchestration tasks (programming and maths).
I add features and tweak it all the time: You just need to get comfortable with spending 3 hours in a chat session every 3 weeks to think hard about how to simplify whatever slop you didn’t simplify the last time you did this :)
Is it open? I'd love to take a look and perhaps steal some ideas :)
Turns out he just used what he knew, TypeScript, and not what best solves the problem. He must have not had his "framework" then.
Steps 1–5 of his framework are increasingly formal ways of saying “figure out what’s going on before doing something,” followed by step 6: “then do it fast”
“let Claude figure out what’s going on before doing something”
“then ask Claude to do it fast with no mistakes”
(edited for accuracy)
> 3. Define the problem
If you do this step well, in my experience, the next steps fall into place quickly.
My version: accurately naming the problem is the essence of problem solving. There are things you usually need to do before you can accurately name the problem, because most problems aren’t presented to you in textbook form.
Once properly defined and framed, it’s obvious how to solve most problems, most of the time.
If you’ve ever worked with someone amazingly good at accurately naming the problem, it’s hard to unsee it. This is one of the hidden talents of the most effective people I’ve met.
It is strange how people who say they are "often wrong", never quite sound like they take that into consideration when delivering opinions or thought pieces - I see it fairly often in tech. Anyways; nothing against Boris, but Anthropic/Claude has certainly been the first company (& product) too annoying for me to use.
That's what it took to finally support Agents.md I guess. Boris had to blog about having made a mistake.
How about you stop taking things so personal instead and let others steer more.
Author here. These are unrelated, just happened to both be the same week.
Unlike me, 56 and yet to be wrong once.
It is a burden all of its own.
> I hope it is interesting or helpful
I think you're wrong. I think it's embarrassing. Though I might be wrong.
It seems strange to write a blog post about being proven wrong, without a single example of being wrong. Especially with all the current Claude.md drama.
Everyone loves being wrong in the abstract - this reads more like a self-congratulation than a genuine reflection.
Heh, I have the opposite... uh "problem".
I am often right.
This has two negative effects;
1) I start to believe my own hype because I am so consistently correct- which blinds me to other ideas if I don't actively take steps to humble myself.
(after all, I won't be right very often if I stop listening, as that's part of why I am right, because I listen to people and have a huge amount of context)
and
2) I am able to quickly come to a conclusion which makes people uneasy, as they think I haven't considered all the points even though I have- and it's also true that people have a bias to inaction unless it's an urgent issue.. this is how things get stuck in committees and meetings and continually get kicked down the road because it's difficult to get everyone together and there's a million reasons to keep deferring meetings.
What I'm trying to say is, don't sit there thinking that "if only I could make correct decisions"- because even if you are quick, correct and decisive: you will ruffle feathers.
(at the risk of sounding arrogant: i'm not saying I'm right about everything, like the red baron, I pick my battles very carefully).
a foolish consistency is the hobgoblin of little minds
i am pretty sure it would just be "wrong," not "meta-wrong," and given the title, i am giving myself the point.
I am not even wrong.
This guy doesn't think deeply about any problem, possibly because he has never had to and probably because he has never wanted to. He would probably make a good residential plumber.
You must not be a home owner because good residential plumbers are definitely out there solving the “hard problems” that techies love to bloviate about every day.
We all have a short time on this planet. This desire to always have urgency, speed, more money is insane.
Humans are in trouble
Boris is a scientist, that's all that has to be said here. Keep up the great work.
“I am often wrong” is directly from how to make friends and influence people.
And yet there's none of this uncertainty or caution with his posts that then go on to have massive ripple effects because of his position at Anthropic and the marketting related to claude.
I personally have a lot of anger and frustration with many people in the ai hypesphere that are just mindlessly frolicking around without a care in the world, happy to speak into the megaphone offered by masses that are in a rat-race to avoid some AI dystopian hellscape that keep getting painted by these thought leaders... and then going "oopsies... I am just human guys... y so mad?!"
I think someone can follow "good practices" with a team and get to a bad place, especially with such a novel product (as AI coding agents).
But taking Claude Code as the product of this style of thinking -- who is Claude Code for? Is it for everyone in the world? Well, if you look at the feature velocity, it seems like the answer is intended to be yes ... Claude Code is trying to solve every problem in software development in the world, all at the same time.
So I question the "user model" here.
Here's another thing that is true about Claude Code: it's among the most inconsistent and buggy pieces of software I've ever encountered.
- You can move the cursor with the mouse in the composer, but not in AskUserQuestion?
- When agents spawn subagents, the model name is inherited from the main agent, and seemingly none of the (4! yes, 4!) subagent tools seem to get this right (except for Explore, which seems to be fixed to a weaker model)
- Sometimes, when my usage limit halts, my agents will pick up when it refreshes (within ~2 hours or something) ... other times, nope -- even within the usage limit?
This is a sampling of my own experiences using this thing frequently. Are these sorts of details not important? Maybe not: I'm not at the level of this team, and may never be.
But I think it's a reflection of agentic engineering ... a somewhat embarrassing one, from my perspective. It paints a picture of a team who can't quite get the details right, even with the assistance of purported extremely powerful AI tools, even internal ones which we don't have access to?
I think when people look back on 2025 -- Boris is going to have his name right there in the books ... Claude Code, coding agents -- Anthropic (& Boris + team) made the first move.
But now it's 2026, and people know how harnesses work, and heavy lies the crown.
Aha, another iterative improvement variant. (see also eg OODA and PDCA)
I'd argue that Boris Cherny is in fact nearly always wrong;
except when he's right, which is in the final pass just before the work is done. ;)
7. Reflect on that goal?
> I love being wrong
I prefer being right, but to each their own.
I like being right so much that, when I'm wrong, I change my mind, despite how embarrassing that is.
Translation: "I'm so wise guys, that I'm using epistemic humility principles to hide it"
Heaps of comments here but nobody has pointed out he defined the goal after defining the problem. That's cheating.
Step one is to define the goal. Then you work out how to reach the goal. (Gather information and form a hypothesis) Then you take steps toward the goal. (Test your hypothesis to make sure you stepped in the right direction.)
Just retire at this point, you don't even have to work.
The most annoying thing about this is that the steps are out of order unless the definition of problems and goals are intermixed 1. define a goal 2. understand information and gaps in information 3. define and prioritize the problems to achieve that goal 4. identify potential solutions for top problems and prioritize 5. measure success
and do all of it with urgency (magically managers want everything done with urgency)
I, I, I, I, I, I, I, Me me me me me me
This is what happens when you let things get to your head. You work on a (highly inefficient, often broken) piece of extremely basic software by prompting a slot machine and hoping it works. Calm down and stop forcing your shitty framework on employees, you're the most replaceable cog of all.
Yeah? A very basic process, most people use some variation of this implicitly without talking about it.
This sounds like a slightly narcissistic manager who thinks people are doing it wrong if they don't act like small copies of him. There are multiple ways to reach the same outcome and a lot depends on your information. E.g. I typically have a mental model of a system in my head, meaning when some problem needs fixing very likely I already know where it would need fixing and already think about the various future implications arising from a different fix. A point that is totally absent from that framework.
Being a senior dev myself I have seen enough good software turn bad to know that seemingly innocent technological decisions can come with huge and lasting implications. The fact that this is missing here is speaking volumes about the lack of experience at display here.
Your task as a manager isn't to create small copies of yourself. Your task is to know each persons weaknesses and strengths and compose the work in such chunks that the weaknesses have little effect, while the strengths multiply. For this you will first of all have to trust your people and lead them to discover certain ideas themselves.
E.g. if you feel someone always jumps gun-ho I to the task without doing the research, just tell them to give you the research first. Do that a few times and they might realize how useful that is.
quick - someone turn this into agentic harness :)
sorry, but when you reach certain levels of influence, you need to slow down, be more methodical, and not be so flippant