Malleable Software

https://www.mdubakov.me/malleable-software-will-eat-the-saas-world/

by tablet • 11 months ago

113 93 11 months ago

knowannoes• 11 months ago

Malleable software will eat you whole.

None of this makes any sense. Do you know how computers work?

This "AI" summer has turned into a drug fueled orgy of magical thinking. I am at my tether's end. I need to leave this industry to preserve my sanity at this point.

misiu1• 11 months ago

where to

fuckaj• 11 months ago

puppy training on a farm

dirtybirdnj• 11 months ago

a drug fueled orgy of magical thinking

This is just "management" thinking. You can thank your local MBA for poisoning the well for everyone else.

Leaders produce value. "Managers" are dumb parrots.

faeyanpiraat• 11 months ago

Rigidity helps in trusting the system.

Malleability / flexibility can introduce unreliability.

We need to get over a hump, where software becomes more humanlike, but just like with good engineers over time we can probably arrive at a place where we can trust our new malleable solutions just like a new colleague turning out to be great.

101008• 11 months ago

A lot of people been saying this lately, that LLMs are going to make SaaS obsolete because you will be able to build the alternative yourself without the need to pay.

But (and I'll copy & paste a comment I wrote a few days ago) I disagree. This existed way before LLM. Open source alternatives to most products are already available. And install them and deploy them is much easier than do it with LLMs, and you get updates, etc.

People don't want the responsability to keep them updated, secured, deployed, etc. Paying a small amount will always be more convenient than to maintain it yourself. The issue was never coding it.

tablet5• 11 months ago

This is not what the article is about. Main idea is that rigid software can finally be replaced by flexible, since flexibility is no longer such expensive

prmph• 11 months ago

Nope, anyone saying this does not understand fundamentally what software is. This so-called malleable software is a recipe for chaos.

bryanrasmussen2• 11 months ago

hey yeah, there's no need to have a payment provider to take care of all your taxes being paid correctly and on time. We have AI!

This would be one of the greatest entertainment events of the 21st century! Shame about all the destruction that will happen as a consequence of course, but ...entertainment!

actionfromafar• 11 months ago

Whole governments run in that mode now.

o_m• 11 months ago

Counter argument: people want simple systems that are easy to update, secure, deploy etc. I've been burned so many times by being an early adopter of a simple product for it to add too many features and shifting focus along the way, leaving the early adopters as second class users. This usually happens because investors wants a return on their investment by enshittifying the product.

As self hosting with Docker and getting help from LLMs gets easier I can totally see a future where more companies self host. Having to deal with SaaS companies also takes a lot of time (licenses, hidden limits you can reach at any time, more complex privacy policy, approval from management), especially as they usually end up selling after a couple of years. The responsibility to self host isn't that bad all things considered.

I don't think we'll see companies vibe code the replacement of their software, but it might help them self host open source alternatives.

cpursley• 11 months ago

What percentage of companies do you think have the technical know how to even fire up their own cloud application and database instance even with all the LLM assistance in the world? Outside of companies in the software space and some of the largest orgs, I’m gonna guess maybe 20%?

firemelt• 11 months ago

not everyone tangled himself with fibery or linear

lmao

malleable software? what a joke

tablet5• 11 months ago

Good article about it btw https://www.inkandswitch.com/essay/malleable-software/

rsav• 11 months ago

Is this just low code all over again, except this time with some nondeterminism thrown in?

s1mplicissimus• 11 months ago

I would put the current LLM coding hype more in the "scaffolding" box, but surely with some nondeterminism thrown in ;)

cyco1303• 11 months ago

For six years I worked in a SaaS startup that built an applicant tracking system (a tool to manage recruitment efforts in big/mid-sized companies) tailored for the local market of the country we lived in. My experience tells me that our main value was in forcing them to rethink their recruitment processes, not adapting to their existing ones that were usually all over the place.

As much as I want to believe the opposite to be true as a “power user”, good tools often force you to adopt better practices, not the other way around.

tablet5• 11 months ago

The problem here is in definition. Context is quite diverse and better practice for team A is an absolute disaster for team B.

cyco1303• 11 months ago

Absolutely. When we started growing (I was employee #3, we were about 20 people when I left), we didn't use our own product for our own needs. It wasn't designed for a tiny startup, it would be like building a sand castle with a bulldozer.

But we started as a "boutique" company that implemented everything requested by our then small number of clients (mainly out of desperation, we were self-funded and we didn't have much leeway, we needed those clients). It was as flexible as it gets before the LLM times.

But after a while, you start noticing patterns, an understanding of what works and what doesn't in a given context. Our later customers rarely requested a feature that we didn't already have or we didn't have a better alternative of. It's not like we had a one-size-fits-all solution that we forced on everyone. We offered a few alternative ways of working that fit different contexts (hiring an airline pilot is a very different context than hiring a flight attendant). And in time, this know-how started to become our most important value proposition.

At some point we even started joking about leaving the software business and offering recruitment consulting services instead.

cyco1303• 11 months ago

In fewer words: It was already a fairly flexible and customizable tool. But then came a time when a client requested faster horses we could show them our car instead and they recognized the value. (And occasionally, when _they_ requested a car instead of our faster horses, _we_ recognized the value and implemented it).

flappyeagle• 11 months ago

They should use different tools then.

Malleable software enables infinite variations of tools when the correct number is in the single digits.

sarchertech2• 11 months ago

I worked for a company that provided a mobile friendly job application form that integrated with major applicant tracking systems (back when they didn’t provide mobile friendly forms).

Our biggest value was getting customers to remove all the extra questions on their applications that had built up over years of management changes that no one had any idea why they were even asking.

lioeters• 11 months ago

good tools often force you to adopt better practices

Just wanted to highlight this excellent statement. It's like having a strict type system that enforces certain rules are always met. It provides consistency and predictability.

rethink their recruitment processes

This context is relevant to the kind of software system that was needed. To improve their processes, it was necessary to impose an explicit top-down order to the existing mess.

Malleable software, on the other hand, feels more suited for personal computing, greenfield projects, or small teams with members working independently as well as collaboratively. Particularly in the early stages of product R&D, strict rules can be a source of friction in the creative process.

Strict better practices and well-designed tools are discovered and developed through open and flexible explorations, as a kind of distillation of knowledge and experience.

mhogers• 11 months ago

data layer > business logic layer > presentation layer

I believe the presentation/analytics layer has become malleable, possibly parts of the business logic layer - you still need a higher % of trustworthiness than LLMs can provide for parts of the business and data layers.

a57721• 11 months ago

you still need a higher % of trustworthiness than LLMs can provide for parts of the business and data layers

For many domain-heavy systems, it's not even the trustworthiness; just getting the business logic right requires a lot of work and lots of iterations with in-house domain experts and clients, there's no way LLMs can do that.

anonzzzies• 11 months ago

For myself yes. I just have claude code running and it's replacing everything i'm doing company and personal wise with custom stuff. However, most people, like my colleagues/employees, want rigidity and do not want to learn new stuff generally. They want to focus on completing their tasks and don't want to have a quicksand saas underneath them. If it helps completing tasks faster then maybe.

Also training new people is annoying when things change too often; people can already use Jira/Linear/Monday/whatever , they don't want some completely flexible thing that is malleable.

Also, people are not all perfectionists with long term goals and visions. People who 'change' some part of their work flow that helps them NOW; they won't care about speed, scaling, deployment etc, so they will do something to make their work easier and then leave it there and possibly ignore it forever to rot. Which might have all kinds of fun implications.

I guess when we have AGI with a few 10 million+ context window for cheap, it will be different but the current llms would just leave a massive amount of rot all over the place, quickly forgotten and not usable by anyone but the original creator.

lagrange77• 11 months ago

This is comparing two orthogonal properties.

SaaS is a business model while malleable vs. rigid is a property of the software itself.

runroader• 11 months ago

Sure, if by SaaS you mean hooking together software that is essentially websites. Major industrial software that costs thousands per seat like Ansys or Dassault are not getting replaced by something that "AI" can cobble together.

The parts of SAP that's composable workflow stuff? Doubt it, because the types of ABAP workflows in SAP that might be "malleable" are the sort of stuff that often legally requires correctness and reproducibility - kinda the exact opposite of a good LLM use-case.

And as much as I'd like to actually own my software, SaaS is preferable for major corporations for lots of legal and accounting reasons like easier revenue recognition. They're going to keep pushing it because it makes all the parts of being a software company that don't include writing the actual software easier.

jaimebuelta• 11 months ago

I see some of this, from the point of view that it's going to be cheaper to create bespoke solutions for problems. And perhaps a "neoSaaS" company is one that, from a very bare bones idea, can create your own implementation.

But, at the same time, there are two issues:

- Companies can be really complex. The "create a system and parametrise it" idea has been done before, and those parametrisation processes are pretty intensive and expensive. And the resulting project is not always to be guaranteed to be correct. Software development is a discovery process. The expensive part is way more in the discovery than in the writing the code.

- The best software around is the one that's opinionated. It doesn't fit all the use cases, but it presents you a way to operate that's consistent and forces you to think and operate in certain way. It guides you how to work and, once going downstream, they are a joy to work with. This requires a consistent product view and enforcing, knowing when to say "no" and what use cases not to cover, as they'll be detrimental from the experience. It's very difficult to create software like that, and trying to fit your use case I'll guarantee it won't happen.

These two things tension any creation of software, and I don't think they'll go away just because we have a magical tool that can code fast.

tablet5• 11 months ago

Your arguments are totally valid, niche tools will be alive and well. I think my take is that even in niche tools we will see a lot of generalization and more flexible niche tools will eventually win.

crote• 11 months ago

The problem is that software can be too flexible. A great example is companies ending up using Excel as a load-bearing database, relying on a bunch of incomprehensible macros to execute critical business logic.

Sure, it's flexible, but are they really better off than a competitor using properly-engineered one-off software? In the end, is there really a difference between software development and flexible-tool-configuration?

ethbr12• 11 months ago

The best software around is the one that's opinionated.

This. And it isn't going to change.

The post avoids trying to answer "Why are opinionated tools popular and effective?"

The answer is that a standardized process that they encourage is often more efficient than whatever bullshit {random company} came up with in-house.

Malleable software needs to produce two equivalently good outcomes to beat opinionated:

1. Improve the underlying process at the customer's business (in terms of effectiveness)

2. Avoid a customization maintenance burden

The seductiveness of "just for you" bespoke solutions is they avoid (1) by telling the customer what they want to hear: you're so brilliant, your process is actually better, our product is a custom fit for your exact process, etc. That's bullshit -- a lot of customer processes are half-baked dumpster fires, and their companies would be better served by following standards.

To (2), I am incredibly skeptical on the long-term tech debt that malleable solutions will impose. What happens when there's a bug in the version only you use? Is that going to be the vendor's priority? Oh, you're supposed to fix it yourself? Congrats... we've just added a requirement that these tools are capable of making random mid-level in-house practitioners as competent as focused dev teams. That's a tall order.

Exhibit A that I'd want a follow-up post to address: SAP.

The above are the reason they realized they were trending in the wrong direction and have been dragging their customer base back to Clean Core.

Walk me through how malleable software would work better for SAP as a product, and I'll begin to believe...

ch4s33• 11 months ago

Highly customizable configuration causes all kinds of problems in healthcare, and EHR customizations have actually killed people.

BinaryIgor• 11 months ago

100%; customization maintenance burden is underrated - it simply costs a lot of time and energy to customize things; often there are better uses of this time, especially in the business context

osigurdson3• 11 months ago

I think the article presents a bit of an odd premise - I can make a mini app in ChatGPT today so by 2035 I can create an entire suite of software needed for a given business. What is the requisite change between what I can do now and in 2035? Presumably it is AGI.

OK, so we are in a digital super intelligence world in 2035. The HR department can now just have a conversation with a chatbot and create software to make them more productive. No more configuring SAP widgets or whatever they do today. The chatbot will be like "hey bro, the process that you want to automate doesn't make any sense: here is a better way. And, by the way, I'm terminating your entire department. I'll take care of it from now on". I mean, get real, in a post DGI world there will be exactly zero office jobs and no SaaS software at all.

glitchc• 11 months ago

It doesn't need to be AGI to build complex software. A human software developer can build a complex software system and perform other complex tasks with the same body (play an instrument, fly an aircraft, etc.). Doing all of that with the same resources is what AGI is needed for. Just software, well I'm sure an LLM can eventually become an expert just like it learnt how to play Go.

ozim• 11 months ago

Odd premise is that AGI will have infinite bandwidth to deal with petty stuff like taking over menial stuff of HR departments.

Current AI barely keeps up with generating funny images people ask from it :)

tikhonj• 11 months ago

The best software around is Emacs. Does that count as "opinionated" in your view?

In some ways it is—Emacs does a lot of things its own way, completely unbothered by mainstream conventions—but, at the same time, it's also totally malleable in the sense of this article. What makes Emacs great is a consistent and coherent conceptual foundation coupled with remarkably flexible code, letting you adjust Emacs to your needs rather than adjusting your needs to Emacs.

Or maybe the best software around is situated software. Software that's built by and for a specific set of people in a specific social context. Situated software is qualitatively different from product software, and it works so well because, again, it gives its users real agency and control. Instead of trying to create software that knows better than its users, we can create software that supports its users in whatever ways works for me. The result is still opinionated, but it's opinionated in a categorically different way from what you're describing.

So perhaps the best mainstream software is Excel.

And, while I don't think they're there now, it seems like LLMs are likely to be the foundation for the next Excel.

skydhash2• 11 months ago

You can either go with simple primitives and a way to combine them (emacs, excel, unix) or simple program that just works (notepad, sumatra,…). Anything else is going to be restrictive in one way or another.

godelski3• 11 months ago

As a vim user I agree with all this. Same is true about why I am terminally terminal. I'm able to adapt the tools to me so that I may get the most use out of them. Sane defaults are great, but there are no settings which are universal. The only solution to this is to let people adjust as needed.

godelski3• 11 months ago

  > Companies can be really complex

I think this is a great argument for flexible code, though it was unclear to me that the author of that post was talking about that.

  > The best software around is the one that's opinionated.

I think I might be on the same page as you but I would say that the best software is written to be an environment more than a specific tool. You're absolutely right that you can't solve all problems.

tikhonj jokingly suggests emacs but even as a vim user I fully agree. Like they say, the beauty of it is that the complexity draws from simpler foundations. It is written as an environment rather than just as a text editor. Being written that way lets it adapt to many different situations and is what has kept both vim and emacs alive and popular after all these years. There's a constant in software development: requirements change with time. The point of writing an environment is that you're able to adapt to these changes. So any time you write a tool that tool is built out of that environment. Anything short of that means the tool won't be able to adapt as time marches on.

I definitely agree that writing software like this is hard but I'm not sure if it is harder. It takes more work up front but I'd argue it takes less work in the long run. It's just that in the long run many efforts are distributed across different people and time. But hey, good flexible code also tends to be much easier to read and that's big short term benefit to anyone coming into a mature project.

hresvelgr• 11 months ago

Inflexibility is the quality I desire more and more as I get older. I don't want to force my software to do dumb shit, I want my software to force me into avoiding dumb shit. This is how you keep a system understandable. That quality is far more precious than the ability to connect complicated enterprise spaghetti machines into each other.

skydhash2• 11 months ago

Stability is something that I also desire more and more. Take my digital clock: I put some AA battery in it, adjust the time, and the only thing I do is look at it (its purpose) for the next months or year. I want my computer to be like that. No surprise when updating.

tonyhart7• 11 months ago

this is also dependent on the field of SaaS fill

good luck replicate this on financial,health,military,cyber etc field

OtherShrezzing2• 11 months ago

I think most of the SaaS stuff that benefits from being malleable already is malleable - just slowly malleable. I can configure a Trello or GitHub Actions in whatever way I want. Meanwhile I really want my email, messenger, or banking apps to be exactly the same every time I use them. I'm not clear how adding a non-deterministic UI or business logic layer is going to fundamentally disrupt or improve experiences like Jira or Visual Studio.

Maybe we're in some kind of local-optimal, where all project management software has coalesced around a few user journeys, and there's some better approach out there to be discovered.. But I don't see why an accounting software company, games studio, or vehicle manufacturer, would dedicate even 1% of its resources into crafting a malleable bespoke project management software toolkit.

It goes against the concept of comparative advantage, and I can't think of any successful enterprise that's bet against comparative advantage and won.

adrianmsmith• 11 months ago

Meanwhile I really want my email, messenger, or banking apps to be exactly the same every time I use them.

I know that's not the point you're making, but I agree with you, alas that's already not the case today, e.g. random device updates nobody asked for, or you log in to your banking website because you need to pay something right now and half the features are gone or different.

marcuschong2• 11 months ago

I agree with you, though I’m not sure I fully understand your point about game studios. That seems like an area where software could evolve in ways that make perfect sense. For example, dynamic worlds, unique missions, unscripted characters, and so forth.

OtherShrezzing2• 11 months ago

My comment is about companies self-serving with malleable & bespoke software. Niantec is unlikely to spend time making a bespoke GenAI version of Jira, just because it's cheaper to do so now than it was before. Every minute they spend making the bespoke project management software is a minute they're not making the next PokemonGo, so they'll pay a third party like Confluence handsomely to produce a predictable project management UI for them.

Games studios have been making dynamic worlds for decades now, and GenAI algorithms are just an evolution of that practice. So I agree that they'll use these tools for their own output, but that output isn't going to disrupt the SaaS business models of companies like Sage/Confluence/Microsoft etc.

random_savv• 11 months ago

This seems like an advertisement for that person's company.

sarchertech2• 11 months ago

Low code/No has alreasy been capable of spitting out CRUD boilerplate forever. Every application turns into a special unique snowflake given time and users.

Code has never been the bottleneck once you’re out of the CRUD boilerplate phase.

endymion-light• 11 months ago

I've experienced this partially, as someone who never uses sheets etc - the capability to create a template that achieves my exact needs within a few seconds is really strong - i can then form malleability in where I want.

But quite honestly, the SaaS world is also due a culling - i'd argue if the software you work on as a SaaS business is replaceable by a malleable piece of AI software, you're closer to Pets.Com than a suitable business model.

damidekronik• 11 months ago

So we need to create tools with very high flexibility, not-so-fantastic onboarding that take a long time to setup and hard to get right - so that AI can fix them? I would rather use Linear (as per example).

dvrj101• 11 months ago

the guy is full of sh*t, his example makes zero sense and looks like badly edited output of creative writing prompt. Not to mention he is the ceo of Fibery so whole thing is just a promotion.

tossandthrow2• 11 months ago

Yes and no.

Malleability is opposed to institution. When everything is hyper malleable everybody will need to be trained and change management will take up a large proportion of time.

The main reason why a lot of people go with Jira is not quality of the software - but the institutional buy-in. Employees, current and prospect, know how to work with the tool.

The greater change is likely labor disruption.

movedx• 11 months ago

Employees, current and prospect, know how to work with the tool.

Just an FYI, this isn't the case at all. I've contracted and consulted at well over 20+ business at this point, and no one knows how-to use that hot garbage.

tossandthrow2• 11 months ago

Let me rephrase that to "be familiar with" then.

Havoc• 11 months ago

It does seem likely that there is a shift towards DIY single use tools.

For anything more than single use there is a bit of a risk of it turning into a mess that collapses under it's own weight. A bit like a couple bad decisions at the start of a coding project comes back to haunt you later. So I struggle to see this replacing core infrastructure of anything in long run.

athrowaway3z• 11 months ago

I'm not sure about software that sells itself on providing a mirror for people to reflect on the structure they use for organization, but I do believe eventually the tools that are going to win are those that do text interface.

Even a visual tool like blender should expose their full GUI as a text API. It needs a bit of adapting, specifically for domain-specific structs (which should be retrievable via calls like 'select-bb' or 'select-coordinates'), but after that's done it's a game changer.

That class of software has a lot of proprietary GUI's that look slick and people are familiar with, but who cares about familiar in a world where the other software lets me point my LLM at its help file and build me whatever gui/tui/script/voice integration I can think of.

hbogert• 11 months ago

The biggest shift LLMs bring to malleable software is moving the focus from designing the solution to defining the problem.

people who know how to formulate the problem already rule the world. AI will just be a catalyst.

If you have good market fit it means you were lucky or you knew which problem your potential customers had.

rco8786• 11 months ago

AI thrives in messy, open-ended spaces

Isn't the exact opposite of what AI is good at? That sounds like a great way to get tons of hallucinations. Every "how to AI good" guide I read is all about providing it with ample structure and narrow instructions.

activitypea• 11 months ago

To be reductive about the example of a malleable task tracker (e.g. Notion), if the LLM correctly interprets the user's prompt, there's about 20-30 useful buttons in the interface, and there's a sequence of those buttons that achieves the user's desired outcome. It's open-ended in terms of what you make, but the set of available actions is fairly tight, and the problem is finding a sequence of those actions.

api• 11 months ago

I’ve been wondering how long it will be before AI can “pirate” SaaS by using it and synthesizing an acceptable replacement. SaaS used to drive real world services would not be amenable to this, but a lot of SaaS is just a UI in front of Postgres. I could see this being possible for anything “thin” like that.

ewf• 11 months ago

Malleable and opinionated aren't mutually exclusive. The best products that solve non-core business processes should be opinionated, that's their value prop.

But yeah, the demand for webhooks, integration options, embedding's will win IMO. I imagine more demand for API-first products.

mariocesar• 11 months ago

I've been in SaaS for about 17 years, and a pattern I've seen is that for companies customers actually love, customization shows up as a milestone. It helps them find their own way of doing things, but eventually they settle into that way.

When it resonates with customers, it feels great.

Customizable SaaS is really just a side effect of the deeper need for control, because with control you can execute better. And when that lines up with capital, market fit, and a clear purpose, it creates more than efficiency, it creates meaning.

The companies that never settle, always chasing new patterns, usually end up with customers who don't enjoy them and employees who don't enjoy the software either.

layer8• 11 months ago

Previous discussion regarding malleable software: https://news.ycombinator.com/item?id=44237881

jonathan-adly• 11 months ago

We have been running this playbook for the last 2 years in healthcare, and we have been super successful. Doubling every quarter over the last year. 70%+ profitability, almost 7 figures of revenue. 100% bootstrapped.

People are still mentally locked in to the world where code was expensive. Code now is extremely cheap. And if it is cheap, then it makes sense that every customer gets their own.

Before - we built factories to give people heavy machinery. Now, we run a 3d printer.

Everyday I thank SV product-led growth cargo cults for telling, sometimes even forcing our competition to not go there.

827a• 11 months ago

I disagree with a lot of the assertions the article makes.

IME: LLMs don't "thrive" in messy open-ended spaces. They handle spaces like that better than traditional code, and traditional code handles structured spaces a lot better, but LLMs still do perform better in structured spaces than unstructured ones. Giving them lists of tools, schemas for data, consistent examples, etc always produces better results than not doing this.

The correlate thesis you have to make is: Do humans work better in unstructured spaces. This can be true for highly creative and individual work, but generally (and especially in the enterprise SaaS world) the opposite is true. The Structure is how you keep the users of the product aligned on one pattern of usage. E.g. giving people the ability to just create whatever ticket fields they want in Linear ends up being useless because you'll end up with 10 fields that do similar things; the friction structure introduces is necessary because while it can cost a few seconds up-front as users learn how things are done, it saves time down the line as your organizational tools (filtering, dashboards, reports, etc) are aligned on the structure.

Its also 100% the case that oftentimes companies buy SaaS tools not to solve their problems, but to help them better structure the problems they have so they're even solvable at all. Think about Sentry: The end-goal of Sentry is to solve issues in web applications, for sure. But that's not why people buy it. I could do that without Sentry; but Sentry adds structure to the errors, it adds deterministic workflows to the errors, and it provides excellent SDKs to report them. Sentry's value isn't actually in the end-goal; its value is in the structure it adds every step before the end-goal. Accelerating the inputs by adding formality to them ends up accelerating the end-goal.

gherkinnn• 11 months ago

So a well-designed and fast saas like Linear will get replaced by a messy solution + AI? Might as well be the case, but what a waste of effort.

smithclay• 11 months ago

My favorite "malleable" example this past week that I recommend: using Gen AI to rewrite, customize, or create dotfiles* for your laptop/desktop.

Defining a bespoke system config was always something that never had the time, desire, or deep familiarity with various config formats to do. With Claude Code or ChatGPT can now generate Wezterm/tmux/neovim/etc config files and get exactly what I want not knowing a single thing about specific config file formats.

* https://dotfiles.github.io

etler• 11 months ago

For the Linear example, I would argue that adding too much flexibility would make the tool worse. Rigid workflows are consistent making them more robust and scalable. They work consistently needing less guidance or intervention.

When you find a rigid workflow that is both widely applicable and useful, that's how the most valuable companies are made. Those workflows can scale up with little intervention giving them incredibly high yield.

I think the new challenge is introducing flexibility in a controlled manner so we can minimize the added inconsistency needed to achieve broader goals. That way the workflow can find the right balance between robustness and flexiblity for the task at hand.

godelski3• 11 months ago

This felt like a very weird read to me. As someone who is very pro writing flexible software.

IME getting LLMs to write flexible code is not easy and one of my biggest turnoffs for coding agents. The code they tend to produce is usually more standardized and lacking of context. It'll "work" but that doesn't mean it "works".

And I think almost everyone is aware of malleable software, it isn't very niche. If you took a programming class your teachers probably told you to write functions and make them small and self contained. A lot like the Unix Philosophy[0].

One thing I find different about this style is that it tends to shift time away from writing lines of code to planning. Which, again, the planning part is what the LLMs tend to be weak at[1]. You can't plan everything but a benefit of flexible code is that you can adapt to new challenges and requirements. It's a good style because a constant in software development is that the requirements by the end of the project (does it end?) are going to be different from those you started with. If your code is inflexible then you have to hack around your existing code. But if it is flexible you add the new routines and leverage existing ones as much as possible. But for that to happen the existing routines should be decently singular in purpose. Function call overhead tends not to be the bottleneck but if the profiler says they are then yeah, fuse functions as need.

[0] https://en.wikipedia.org/wiki/Unix_philosophy

[1] I find their strengths are in writing generic routines, helping narrow in on fuzzy or difficult to describe things that are hard to google, helping you get up to speed on a novel codebase, ditto for a library or language, or simply just throw away code. But not for things where foresight is necessary. But yes great when your job is to glue things together. It's just that if you're doing a lot of gluing there's probably a better way

snowstormsun• 11 months ago

Now, in many cases, LLMs can handle the "how" for you. You describe what you want in plain language, and the system works like a programmer or system analyst: breaking your problem into building blocks, mapping a flow to solve it, and creating the first version.

People keep saying things like this, but what's an actual example of a viable product created that way? Talk is cheap, I want to see it.

Malleable Software

Related Stories