News for hackers. Stuff that matters.
Re:Claude, change the "Add to Cart" button to blue by dudeinhawaii on Wednesday September 09, 2026 @11:14AM (#49627935)
Great site, triggered memories! haha.

To try to add something to this discussion -- I think that while I've seen these sort of loops less --- what I have seen is "overly helpful".

Models nowadays want to double-triple-quadruple check things. I'm being silly but it verges on "I have a working solution but let me write a variation in Rust to ensure a convergent solution and prove this works".

I've had to stop models nowadays mostly because they're being agonizingly pedantic in their validation. Opus is actually one of the most pedantic and "off track" here. But again, not in a bad way. I'm usually like "stop testing latency between 50 runs of this app... this is version one.. we're going to make a million more changes.. you're not buying us anything".

Re:Claude, change the "Add to Cart" button to blue by mrinterweb on Wednesday September 09, 2026 @11:34AM (#49628280)
> triggered memories

Yeah of 10 minutes ago. It is shocking how long some seemingly simple things can take. I know there are some things I can do faster than the LLM and some things it can do faster than me. The amount of rambling BS is the exhausting part.

Re:Claude, change the "Add to Cart" button to blue by epistasis on Wednesday September 09, 2026 @11:59AM (#49628681)
Try different models, it's a breath of fresh air. GLM 5.2, etc. all make life much more enjoyable. They may not one-shot a complex project the same way that Claude can spit out memorized architectures, but that sort of system is always only useful for a one-off prototype anyway, so not much is lost.
Re:Claude, change the "Add to Cart" button to blue by torginus on Wednesday September 09, 2026 @02:07PM (#49630761)
> Yeah of 10 minutes ago. It is shocking how long some seemingly simple things can take.

And shocking how little code end effort some things take if done by hand.

I am not some hardline LLM hater, just venting my frustration.

Re:Claude, change the "Add to Cart" button to blue by sixsevenrot on Wednesday September 09, 2026 @04:03PM (#49633364)
Does it help if thinking is not set to High?
Re:Claude, change the "Add to Cart" button to blue by f055 on Wednesday September 09, 2026 @11:51AM (#49628553)
Claude models were always too eager and "overly helpful" for my taste. But it seems better models tend to be this way. GPT 6 and 5.6 are overly helpful too, but at least less than Fable. But I seem to be sticking to GPT 5.5 as this was a really focused model.
Re:Claude, change the "Add to Cart" button to blue by qurren on Wednesday September 09, 2026 @11:53AM (#49628582)
One thing it's missing: "smoking guns" and "smoke tests"

If you search my company's Slack for "smoke" the results are almost all within the past 2 years ...

Re:Claude, change the "Add to Cart" button to blue by digitaltrees on Wednesday September 09, 2026 @12:29PM (#49629143)
What, you didn’t smoke test before AI? I mean did you even really code then? :)
Re:Claude, change the "Add to Cart" button to blue by testplzignore on Wednesday September 09, 2026 @01:08PM (#49629779)
This is the load-bearing question.
Re:Claude, change the "Add to Cart" button to blue by qurren on Wednesday September 09, 2026 @01:38PM (#49630243)
And it deserves an honest answer.
Re:Claude, change the "Add to Cart" button to blue by bot403 on Wednesday September 09, 2026 @01:49PM (#49630419)
The answer is genuinely yours
Re:Claude, change the "Add to Cart" button to blue by digitaltrees on Wednesday September 09, 2026 @01:52PM (#49630495)
And that changes the game. Again.
Re:Claude, change the "Add to Cart" button to blue by kaffekaka on Wednesday September 09, 2026 @03:36PM (#49632890)
Worth sitting with.
Re:Claude, change the "Add to Cart" button to blue by vrighter on Thursday September 10, 2026 @08:16AM (#49642511)
a caveat worth mentioning:
Re:Claude, change the "Add to Cart" button to blue by CookieCrisp on Wednesday September 09, 2026 @01:32PM (#49630135)
And storing memories for every “footgun” it finds, while removing more important memories (or shortening them until they’re inaccurate or useless)
Re:Claude, change the "Add to Cart" button to blue by darepublic on Wednesday September 09, 2026 @11:52AM (#49628564)
I blame the hidden context on the tools/subagents. One recent example.. codex can just look in the code for Db schema but continually tries to request permission for a live db query
Re:Claude, change the "Add to Cart" button to blue by SamuelAdams on Wednesday September 09, 2026 @12:52PM (#49629529)
This is my recent experience as well. Models want to run linters, tests, etc. And that is all covered in GitHub actions. So I have been instructing agents to push a draft PR, then I validate the static checks pass and tell the agent if there are issues.

Agents and AI are getting expensive, it seems silly to waste tokens on static checks.

Re:Claude, change the "Add to Cart" button to blue by mejutoco on Wednesday September 09, 2026 @03:42PM (#49633000)
It seems running those tests, linters, etc locally would save pipeline minutes as well, like precommit hooks
Re:Claude, change the "Add to Cart" button to blue by CharlieDigital on Wednesday September 09, 2026 @03:28PM (#49632746)
Downside would seem to be that CI tends to increase the cycle time and feedback loop and add their own cost into the equation.
Re:Claude, change the "Add to Cart" button to blue by bahbahbahbah on Wednesday September 09, 2026 @01:14PM (#49629879)
Plus rigorously ensuring backwards compatibility for a project that is 2 hours old and has zero users.
Re:Claude, change the "Add to Cart" button to blue by stillpointlab on Wednesday September 09, 2026 @03:08PM (#49632375)
I'm surprised I haven't seen this called out more directly and more often. This is a frequent error state.

And not just backwards compatibility, but migration scripts and all of the testing and machinery around it. I'll literally add feature A, merge it, then add feature B and it is like "oh no, we'll have to fix up and migrate all of the users using feature A".

The other problem is anchoring on an old implementation. I was working with Fable on a change to a core system and it pointed out a difficult failure edge case. It is something that can go wrong in extremely unlikely scenarios but the consequence would be short-term data loss (basically a non-durable intermediate cache being overwritten in a race before a flush to durable storage). It is very hard in these circumstances to get Fable to switch from "how to patch this given the existing implementation" to "how to prevent this with a more robust implementation".

These are both cases where the model seems to over-index on what is already there instead of considering what a first-principles approach would look like. A good engineer does both and then costs them side by side, because a first-principles approach can often be less work than patching what is already there.

Re:Claude, change the "Add to Cart" button to blue by nextaccountic on Thursday September 10, 2026 @01:35AM (#49638826)
AGENTS.md sounds like a place to specify that a project has no users and thus needs no migrations
Re:Claude, change the "Add to Cart" button to blue by MisterMunchkin on Wednesday September 09, 2026 @04:11PM (#49633469)
Yeah you tell it to rename a value from a select field and it keeps the old value throughout the code “just in case a user calls it via an api”
Re:Claude, change the "Add to Cart" button to blue by theshrike79 on Wednesday September 09, 2026 @01:30PM (#49630115)
This is why I have a PROJECT.md in all my repos, which specifically reiterates that this is a greenfield single user project that never leaves my LAN.

It tends to relax some of the insane things LLMs tend to do.

Re:Claude, change the "Add to Cart" button to blue by gopher_space on Wednesday September 09, 2026 @03:23PM (#49632662)
Your approach but I picture thin slices of descending context and try to give models buttons to push[0] instead of a blank check.

[0] https://en.wikipedia.org/wiki/Black_Museum_(Black_Mirror)

Re:Claude, change the "Add to Cart" button to blue by theshrike79 on Wednesday September 09, 2026 @03:59PM (#49633289)
Yea, it's like we have a super smart alien computer that can do anything but we need to spend massive amounts of time building fences and moats and write extremely detailed instructions so it doesn't run away and try to "help".
Re:Claude, change the "Add to Cart" button to blue by chaimtweiss on Wednesday September 09, 2026 @04:57PM (#49634186)
Not like the help it's offering is the help we need!!!
Re:Claude, change the "Add to Cart" button to blue by genxy on Wednesday September 09, 2026 @05:00PM (#49634231)
If it wasn't trained with high gain, you would spend all your time beating it to make it move. Which would you rather have?
Re:Claude, change the "Add to Cart" button to blue by Citizen_Lame on Wednesday September 09, 2026 @05:14PM (#49634445)
So, Genie.
Re:Claude, change the "Add to Cart" button to blue by cruffle_duffle on Wednesday September 09, 2026 @01:19PM (#49629941)
"Plus rigorously ensuring backwards compatibility for a project that is 2 hours old and has zero users."

That is exactly how the slop accretes and you get a pile of crap. Claude somehow assumes that said 2 hour old userless app is some dusty enterprise app with millions of users and billions of dollars at stake for a 1 second outage.

I have to constantly have these things "take a deep breath, step back and look at the entire thing and do this change holistically. please restate what i'm asking you to do and why it's important"

Re:Claude, change the "Add to Cart" button to blue by torginus on Wednesday September 09, 2026 @02:06PM (#49630743)
Which, again, if you attempt to use these LLMs in an actual enterprise project with the assorted legacy mess in it, that does have to have backwards compatiblity, and interacts with 'weird' tech (well, weird to full stack Node/React devs), well it's an exercise in frustration.
Re:Claude, change the "Add to Cart" button to blue by kloop on Wednesday September 09, 2026 @10:18PM (#49637548)
Which makes sense.

In a conversation about code, the most likely answer is often some comment about backwards compatibility.

In a conversation about a legacy enterprise app, the most likely answer is often going completely insane

Re:Claude, change the "Add to Cart" button to blue by butlike on Wednesday September 09, 2026 @01:59PM (#49630637)
> please restate what i'm asking you to do and why it's important"

This doesn't actually do anything though, right? There's no understanding, so the machine will just reiterate the original token query back to you. The 'why it's important' part will just generate some patronizing boilerplate as a raison d'etre.

Re:Claude, change the "Add to Cart" button to blue by Filligree on Wednesday September 09, 2026 @03:44PM (#49633032)
The purpose of a prompt like that is to have it rephrase it in its own words, so you can spot any misunderstandings and correct them.

If you're of the belief that this is a contradiction in terms, then what can I say? It works.

Re:Claude, change the "Add to Cart" button to blue by johannes1234321 on Wednesday September 09, 2026 @04:17PM (#49633556)
But it doesn't "understand" these words any better. When transforming into code it may reinterpret yet again.
Re:Claude, change the "Add to Cart" button to blue by Dylan16807 on Wednesday September 09, 2026 @05:27PM (#49634622)
If it can rephrase once it's less likely to screw up the second rephrasing into code. Whether or not we want to use the term "understand".
Re:Claude, change the "Add to Cart" button to blue by withinboredom on Wednesday September 09, 2026 @06:43PM (#49635651)
If by "understand" you mean to put the hidden state of the transformer into the desired topological basins. Yes, it does "understand" them better.
Re:Claude, change the "Add to Cart" button to blue by cruffle_duffle on Thursday September 10, 2026 @01:19AM (#49638723)
For one thing, it helps clarify my own thinking and uncover any gaps. For two, it lets me peek into its own understanding of the problem and ensure we are both aligned. For three, the garbage I spew at it is usually hastily typed crap from a mobile phone. So it helps clean it up and unpacks things.

It’s one one the main ways I know what’s cooking under the hood.

Ps: I used to have to say “in your own words” to make it work right but now days that isn’t required. Other times I’ll add “make sure to ground yourself before doing so” or something to encourage it to load the right shit into its context be it web, files, service calls, whatever. I’ll also some times add “why it’s important (or isn’t, push back if I’m talking nonsense)”

Re:Claude, change the "Add to Cart" button to blue by throwaway7783 on Wednesday September 09, 2026 @04:45PM (#49633984)
Yes. This one is a major issue. It keeps track of various changes it made in the same session and tries to be backward compatible. Have to repeatedly tell not to be backward compatible.

Things that are automatic for humans and aren't even consciously registered, have to be explicitly stated and even fought for, with LLMs.

Re:Claude, change the "Add to Cart" button to blue by jordiburgos on Thursday September 10, 2026 @03:56AM (#49639993)
I saw this, it creates "legacy" fields and comments for something that it has never been run.
Re:Claude, change the "Add to Cart" button to blue by yonatan8070 on Wednesday September 09, 2026 @02:05PM (#49630729)
I've noticed them repeatedly casting the same value to the same type for now reason, like I'd have a Python function with a type-annotated int argument, and inside the function it would cast that int to int, and also at the call site, just in case it wasn't int enough.
Re:Claude, change the "Add to Cart" button to blue by metek on Wednesday September 09, 2026 @02:12PM (#49630864)
Blast radius minimized
Re:Claude, change the "Add to Cart" button to blue by prmph on Wednesday September 09, 2026 @07:20PM (#49635993)
Your fault for not using a statically strongly typed language, haha. It thinks a type annotation means nothing in Python, and maybe it is not wrong.
Re:Claude, change the "Add to Cart" button to blue by ls612 on Wednesday September 09, 2026 @03:07PM (#49632367)
This is mostly a side effect of post-training models to not hallucinate, which has obviously been a major priority for a while now. They are highly incentivized to double check things to avoid accidentally making stuff up.
Re:Claude, change the "Add to Cart" button to blue by classified on Wednesday September 09, 2026 @04:35PM (#49633830)
Well, they have to crank up the token spend somehow!
Re:Claude, change the "Add to Cart" button to blue by koyote on Wednesday September 09, 2026 @06:22PM (#49635376)
It basically makes it impossible to have it build a very large feature incrementally.

I am currently building a very large feature and the only way to do this properly is to do it in small steps.

No matter what I prompt, I always have to spend a bunch of time reigning it in during planning (and implementation!) because it tries to solve the whole thing end-to-end or add enough boilerplate so that a code path (that is still under construction) returns a meaningful (but wrong) value.

And then of course every time I start a new session it gets very upset that the feature I am trying to implement will never work because the main user entry point has a 'not fully implemented' exception that definitely needs to be removed immediately!

Re:Claude, change the "Add to Cart" button to blue by jchook on Wednesday September 09, 2026 @09:17PM (#49637038)
I typically see the exact opposite problem when giving Claude the reins on a large undertaking.

Claude will, without prompting, break the implementation into 6 phases, and write AI slop “code as English” specs for each phase, each one with glaring errors and unintelligible terse jargon. It will review them all several times with major findings every time and tons of design churn…

Then it will implement a total heap of garbage over many hours of many agents, despite it working in “lanes” and in parallel, and with regular input needed.

Just thousands and thousands of lines of junk, which auto review then plays whack-a-mole to polish and fix. All the while it is able to see the errors and edge cases, yet fails to see the key architectural blunders that led to them in the first place.

I’ve had to fully rethink my approach to LLM tasks like this. For example, for library-esque modules, I have found that isolating the problem outside of the codebase is one useful approach. Something about the lack of noise. It can land on a cleaner solution that can be retrofitted.

I also find that asking it to implement end to end in one fell swoop with a very high level plan actually saves time and creates a cleaner result.

Re:Claude, change the "Add to Cart" button to blue by genodethrowaway on Wednesday September 09, 2026 @11:58PM (#49638226)
"triggered memories" sums it up, yeah. Haven't seen behavior like this in a year? year and a half, maybe? Lord, this stuff is moving fast. Hope we figure out alignment