The question is usually understood to mean "on an ongoing basis" and it's essentially lying. The system may not be correct or kind, but I'd have a lot of trouble trusting someone whose response to "the system isn't fair" is "I'll lie to get around it and achieve the outcome I believe is right" enough to hire him.
I could see that, but given that jobs are all at-will it's hard to argue the employer would have any damages for a contract that can be terminated with literally zero notice.
Most employment is at will, but salaried employees usually have specified severance in case of termination for convenience. Misrepresenting qualifications does in fact create a case for damages unless the employee is literally perfect. But civil aside, and ethics aside, it is fraud and could conceivably be prosecuted accordingly.
I personally wouldn't have a lot of faith in the abilities and judgement of someone who doesn't leverage an unfair system to their advantage despite there being basically no downsides to doing so unless im hiring a middle manager to sort through excessive bureaucratic bloat.
This is a bit of a game theory problem. "Training senior engineers" is an expensive and thankless task: you bear essentially all the cost, and most of the total benefit accrues to others as a positive externality. Griping at companies that they should undertake to provide this positive externality isn't really a constructive solution.
I think some people are betting on the fact that AI can replace junior devs in 2-5 years and seniors in 10-20, when the old ones are largely gone. But that's sort of beside the point as far as most corporate decision-making.
This hyper-fixation on replacing engineers in writing code is hilarious, and dangerous, to me. Many people, even in tech companies, have no idea how software is built, maintained, and run.
I think instead we should focus on getting rid of managers and product owners.
The real judge will be survivorship bias and as a betting man, I might think product owners are the ones with the entrepreneurial spirit to make it to the other side.
Product owners and project managers have the soft skills to convince the company that they aren't a drain on its resources regardless of what they actually are.
I've worked for a company which turned from startup to this. Product owners had no clue what they own. And no brain capacity to suggest something useful. They were just taken from the street at best, most likely had relatives' helping hands. In a couple of years company probably tripled manages headcount. It didn't help.
The people who will come out the other side are domain focused people with the engineering chops to understand the system end to end, and the customer skills to understand what needs to be built.
Don't forget the very important role of managing the problem solvers--if you just let the problem solvers run amuck all sorts of problems might be solved.
Yeah, if places like RAND or Xerox PARC or the OG Skunkworks, or even Manhattan Project and Apollo Program taught us, is that you cannot let engineers and domain experts run the show, because if you do, they start doing some world-upending shit like putting GUIs on the Moon, or building nukes, or supersonic jets, or inventing new materials that violate the natural order of things, or they generally just rock the boat too much, continuously disrupting the corporate and political pecking order.
Nah, you have to put them in hamster wheels so they keep generating steady value for the shareholders, and put those in open plan offices so they get too mentally exhausted and distracted to try and change things. Throw in free cheese during good economy to keep them happy, but that's strictly optional.
Your architect, principal engineer etc. (one spot-on job title I've seen is "product architect"), who in turn talks to the senior management.
Basically an engineer with a talent and experience for building products rather than a manager with superficial understanding of engineering. I think the most ambitious teams have someone like this on top - or at least around
I've had your type of product owner, but I've also had a product owner that was an ex-staff engineer. Companies should hire ex-engineer product owners, not strictly people-manager product owners.
Technical background doesn't always help in my experience - it's just a different role. Creating great product requires deep technical expertise to understand where the cutting edge is, vision to understand how it can be expanded and business expertise to understand what makes sense economically. It's just not a manager's job, you can't perform it by collecting customer requirements in a spreadsheet.
If you have 10PMs and 90 devs today, and go to 8 "hybrid" PMs + 2 specialized PMs, you're probably still creating backlog items faster than that team can close them.
So you end up with some choices:
* do you move at the same speed, with fewer people?
* do you try to move faster, with less of a reduction in people? this could be trickier than it sounds because if the frequency of changes increases the frequency of unintended consequences likely does too, so your team will have to spend time reacting to that
I think the companies that win will be the second batch. It's what happens today, basically, but today you have to convince VCs or the public market to give you a bunch of more money to hire to 10x the team size. Getting a (one-off?) chance to do that through tooling improvements is a big gift, wasting it on reducing costs instead of increasing growth could be risky.
With Agentic RL training and sufficient data, AI operating at the level of average senior engineers should become plausible in a couple to a few years.
Top-tier engineers who integrate a deep understanding of business and user needs into technical design will likely be safe until we get full-fledged AGI.
On the other hand I’m pretry sure you will need senior engineers not only for designing but debugging. You don’t want to hit a wall when your Agentic coder hits a bug that it just won’t fix.
There’s a recent article with experiments suggesting LLMs are better at bug fixing than coding, iirc. It’s from a company with a relevant product though.
Case 1.1: AGI soon, you don't need juniors or seniors besides a very few. You cost yourself a ton of money that competitors can reinvest into R&D, use to undercut your prices, or return to keep their investors happy.
Case 1.2: No AGI. Wages rise, a lot. You must remain in line with that to avoid losing those engineers you trained.
Case 2: You quit training juniors and let AI do the work.
Case 2.1: AGI soon, you have saved yourself a bundle of cash and remain mostly in in line with the market.
Case 2.2: no AGI, you are in the same bidding war for talent as everyone else, the same place you'd have been were you to have spent all that cash to train engineers. You now have a juicier balance sheet with which to enter this bidding war.
The only way out of this, you can probably see, is some sort of external co-ordination, as is the case with most of these situations. The high-EV move is to quit training juniors, by a mile, independently of whether AI can replace senior devs in a decade.
Case 1.3: No AGI, tools increase productivity a lot, you have a bigger team and you make them more productive. In the meantime, while everyone else was scared of hiring, you got a bunch of stuff done to gain a lead in the market.
You get high EV because everyone else in your market voluntarily slowing down is a gift-wrapped miracle for you.
(Even in an AGI-soon case - you spent a bit more (let's be serious here, we're not talking about spending our entire bankroll on 18months of new hires here) in short term to get ahead, then you shift people around or lay them off. Your competitors invested that money into R&D? What does that even mean if it didn't involve hiring and AGI happens soon anyway?)
----
(Case 3: AGI soon, you don't need yourself anymore - it's hard to imagine a sufficiently advanced "AGI" that someone only replaces software devs but leaves the structure, management, and MBA-trappings of modern exchange and businesses alone.)
> The only way out of this, you can probably see, is some sort of external co-ordination, as is the case with most of these situations.
You lack imagination. You can eg just charge juniors for the training.
Either directly (which won't really work, because juniors almost by definition don't have a lot of money), or via a bond that they have to pay back iff they jump ship before a set number of years.
Have a look at how airlines and airline pilots pay for their expensive education.
You’re looking at it from the point of view of an individual company. I’m seeing it as a risk for the entire industry.
Senior engineers are already very well paid. Wages rising a lot from where they already are, while companies compete for a few people, and those who can’t afford it need to lean on AI or wait 10+ years for someone to develop with equivalent expertise… all of this sounds bad for the industry. It’s only good for the few senior engineers that are about to retire, and the few who went out of their way to not use AI and acquire actual skills.
An interesting thing to consider is that Codex might get people to be better at delegating, which might improve the effectiveness of hiring junior engineers. Because the senior engineers will have better skills at delegating, leading to a more effective collaboration.
I’m curious about other aspects of this:
- leverage of countries who can host such AI over countries who can’t, will there be a point when countries can’t allow themselves not to have access to „emergency” talent in case they can’t use AI? Recent „choose european”, tariffs show that much of the high end stuff is concentrated in US and China.
- outages happen, does the company stop because the cloud is not working?
- highly regulated companies still can’t use copilot to its fullest because of „can’t show answer because it’s matching public code”
- is replacing all talent safe - in terms of operational or national safety?
I believe cursor now supports parallel tasks, no? I haven't done much with it personally but I have buddies who have.
If you want one idiot's perspective, please hyper-focus on model quality. The barrier right now is not tooling, it's the fact that models are not good enough for a large amount of work. More importantly, they're still closer to interns than junior devs: you must give them a ton of guidance, constant feedback, and a very stern eye for them to do even pretty simple tasks.
I'd like to see something with an o1-preview/pro level of quality that isn't insanely expensive, particularly since a lot of programming isn't about syntax (which most SotA modls have down pat) but about understanding the underlying concepts, an area in which they remain weak.
Atp I really don't care if the tooling sucks. Just give me really, really good mdoels that don't cost a kidney.
I am a major hater of many (most?) crypto applications, which should tell you something since the idea of a decentralized currency outside state control is one that deeply appeals to my principles.
But this is one of the better applications I've seen. Running centralized infra for this specific case is extremely difficult and, generally speaking, it makes sense to give people the option to express to willingness to pay for what's essentially a priority request.
This isn't pay-for-access, it's "I'll offer some reward for you to get the paper now, after which it is still accessible to everyone."
My big quibble is with the implementation: there really doesn't need to be a sci-hub memecoin. Monero is purpose-built for this sort of thing. Use Monero (or zcash, I suppose.) Easy litmus test: if a DNM opened up that only supported transactions in its own "memecoin", how many people would take it seriously? Zero.
Not OP, but presumably it's because it could cement a permanent divide between classes. We still have quite a bit of upward mobility in the US, but health is a tremendous predictor of future outcomes, so gating that to the rich is dangerous to the stability of society in that way.
This seems like more of an issue with accessibility of the treatment than the treatment itself
If we could make most children smart, productive, ambitious, courteous, civil, conscientious, honorable, strong... the value to society is probably high enough to justify covering it for almost anyone.
The society already can invest a lot (through public education) to “make most children smart, productive, ambitious …”.
Somehow society (or indeed parts of it) decided to use it as a tool of further segregation rather than overall prosperity. I’m afraid same might apply to this.
We "invest" more than almost anyone. 38% higher than the OECD average. I don't find discussions about throwing more money at the problem to be constructive so much as a way to ignore other issues at play.
I don't really see how this affects e.g. what I do for my children. I will absolutely be turning them into the closest to superhuman the current state of treatments lets me, traveling internationally if I need to. If someone else decides to segregate access to treatment, that is a separate, wrong act that will not hold me back from giving my children every advantage possible.
(Yes, I understand this is a positional arms race, but 1. that doesn't change the individually-optimal outcome, and 2. that doesn't change that society net benefits from it.)
I don't mean to invest as to spend more money, rather to spend money better and in a more equal way. While USA spends a lot of money on education I don't think it translates in better education on average. Even if this was beneficial for the society in general.
I am, afraid, that this kind of genome modification will further increase divide in a society and turn social lifts off even more. I.e. it's not gonna be your kid to get "improve" brain genes first, and later your kid wouldn't get a chance to get it ever again for their children.
Just to be clear I'm not against of the progress, this thing is fascinating and really shows how awesome humans are. And I get why you'll get it if possible for your kid. I'm just not sure its benefits for the society mean it's gonna be anyhow affordable for regular people.
You will have a really hard time convincing Americans to keep paying high taxes while funding is pulled from their children’s schools and redistributed to inner cities and ruralia. My observations suggest the problem for the latter isn’t financial.
I'm explaining that gene modification will not be considered illegal or bad because the rich will have a vested interest in it being legal. This is a reply to GP saying:
> use legal mechanisms to discriminate and persecute people who are genetically modified
I believe there is no way this will happen, because legal mechanisms are driven by the whims of the rich, and they will want gene editing to be legal. So there will beno legal mechanisms to discriminate against those who have been edited.
I've found sonnet-3.7 to be incredibly inconsistent. It can do very well but has a strong tendency to get off-track and run off and do weird things.
3.5 is better for this, ime. I hooked claude desktop up to an MCP server to fake claude-code less the extortionate pricing and it works decently. I've been trying to apply it for rust work; it's not great yet (still doesn't really seem to "understand" rust's concepts) but can do some stuff if you make it `cargo check` after each change and stop it if it doesn't.
I expect something like o3-high is the best out there (aider leaderboards support this) either alone or in combination with 4.1, but tbh that's out of my price range. And frankly, I can't mentally get past paying a very high price for an LLM response that may or may not be useful; it leaves me incredibly resentful as a customer that your model can fail the task, requiring multiple "re-rolls", and you're passing that marginal cost to me.
I am avoiding the cost of API access by using the chat/ui instead, in my case Google Gemini 2.5 Pro with the high token window. Repomix a whole repo. Paste it in with a standard prompt saying "return full source" (it tends to not follow this instruction after a few back and forths) and then apply the result back on top of the repo (vibe coded https://github.com/radekstepan/apply-llm-changes to help me with that).
Else yeah, $5 spent on Cline with Claude 3.7 and instead of fixing my tests, I end up with if/else statements in the source code to make the tests pass.
I decided to experiment with Claude Code this month. The other day it decided the best way to fix the spec was to add a conditional to the test that causes it to return true before getting to the thing that was actually supposed to be tested.
I’m finding it useful for really tedious stuff like doing complex, multi step terminal operations. For the coding… it’s not been great.
I’ve had this in different ways many times. Like instead of resolving the underlying issue for an exception, it just suggests catching the exception and keep going
It also depends a lot on the mix of model and type of code and libraries involved. Even in different days the models seem to be more or less capable (I’m assuming they get throttled internally - this is very noticeable sometimes in how they try to save on output tokens and summarize the code responses as much as possible, at least in the chat/non-api interfaces)
Cool tool. What format does it expect from the model?
I’ve been looking for something that can take “bare diffs” (unified diffs without line numbers), from the clipboard and then apply them directly on a buffer (an open file in vscode)
None of the paste diff extension for vscode work, as they expect a full unified diff/patch
I also tried a google-developed patch tool, but also wasn’t very good at taking in the bare diffs, and def couldn’t do clipboard
Markdown format with a comment saying what the file path is. So:
This is src/components/Foo.tsx
```tsx
// code goes here
```
OR
```tsx
// src/components/Foo.tsx
// code goes here
```
These seem to work the best.
I tried diff syntax, but Gemini 2.5 just produced way too many bugs.
I also tried using regex and creating an AST of the markdown doc and going from there, but ultimately settled on calling gpt-4.1-mini-2025-04-14 with the beginning of the code block (```) and 3 lines before and 3 lines after the beginning of the code block.
It's fast/cheap enough to work.
I seem to be alone in this but the only methods truly good at coding are slow heavy test time compute models.
o1-pro and o1-preview are the only models I've ever used that can reliably update and work with 1000 LOC without error.
I don't let o3 write any code unless it's very small. Any "cheap" model will hallucinate or fail massively when pushed.
One good tip I've done lately. Remove all comments in your code before passing or using LLMs, don't let LLM generated comments persist under any circumstance.
Interesting. I've never tested o1-pro because it's insanely expensive but preview seemed to do okay.
I wouldn't be shocked if huge, expensive-to-run models performed better and if all the "optimized" versions were actually labs trying to ram cheaper bullshit down everyone's throat. Basically chinesium for LLMs; you can afford them but it's not worth it. I remember someone saying o1 was, what, 200B dense? I might be misremembering.
I'm positive they are pushing users to cheaper models due to cost. o1-pro is now in a sub menu for pro users and labled legacy. The big inference methods must be stupidly expensive.
o1-preview was and possibly still is the most powerful model they ever released. I only switched to pro for coding after months of them improving it and my api bill getting a bit crazy (like 0.50$ per question).
I don't think paramater count matters anymore. I think the only thing that matters is how much compute a vendor will give you per question.
I never have LLMs work on 1000 LOC. I don't think that's the value-add. Instead I use it a the function and class level to accelerate my work. The thought of having any agent human or computer run amok in my code makes me uncomfortable. At the end of the day I'm still accountable for the work, and I have to read and comprehend everything. If do it piecewise I it makes tracking the work easier.
Big test time compute LLMs can easily handle 1k depending on logic density and prompt densitity.
Never an agent, every independent step an LLM takes is dangerous. My method is much more about taking the largest and safest single step at a time possible. If it can't do it in one step I narrow down until it can.
I've been using Mistral Medium 3 last couple of days, and I'm honestly surprised at how good it is. Highly recommend giving it a try if you haven't, especially if you are trying to reduce costs. I've basically switched from Claude to Mistral and honestly prefer it even if costs were equal.
Phone numbers are used like this because in the Year of our Lord 2025, they’re the best way to semi-solve the Sybil problem even somewhat without having to literally do some kind of KYC
This is probably compliance-related. For me, TOTP isn’t “something I have”, it’s another thing I toss into my password manager and sync to all devices.
I really agree with it, but that’s probably their rationale.
Banks didn't support TOTP long before we were able to easily sync them across devices. It's likely more along the lines of banks generally have bad IT departments and outdated digital security policies.
That same rationale wouldn't support SMS as "something I have." iMessage and other solutions easily spread SMS into cloud and PC lands (ones that are more easily accessible than password managers.) More likely it's because of legacy and "good enough" reasons.
Personally I don't put TOTP tokens into my password manager and keep a dedicated app for it, just in case my password manager is pwned.
I'm not really defending it, I'm explaining the mentality. iMessage is probably closer to "something I have" but yeah, often not true for many American users.
I'd probably keep a TOTP app if I actually brought my cell with my everywhere but I really don't feel like it; if I'm heading to a cafe to work for a bit I might need to access something and can't be bothered to bring two devices.
Plus, people increasingly access stuff from cell phones, so it's not a guarantee of "something you have" anymore. And no shot we're convincing everyone to start carrying some kind of hardware token.
You have to remember that cybersecurity is driven by what is secure so much as what is compliant, and increasingly so.
I do the same, and it somewhat defeats the spirit of 2FA, but I still believe it's more secure. It's basically a second password where intercepting it in transit once isn't enough to be able to repeat the login in the future.
My bank sends me 2FA codes in their app, which I then have to type into... their app. No kidding. Both the key and the validation in the same place, really ridiculous. Even something as crap as SMS 2FA would be better. TOTP or FIDO2 would be miles better.
The point here is to avoid the oracles of the world getting everything. Please trust me on this one: fedramp isn’t really about security. Compliance with it doesn’t ensure it and noncompliance doesn’t preclude it. It was okay when originally released but that’s not been the case for years now.
