Good, I will with great pleasure now reiterate my point about people now producing their own code, even complex stuff, rather than downloading potentially malicious and foreign code. Which as a tendency threatens ALL clumsy big ERP service providers selling you SAAS.
Go ahead - I'm ready to be down-voted again and again until folks realize it is inevitable, as is inevitable that many companies in the area of business software are going down down down.
I'm the main dev in a small IT company, my backlog is filled with requests, it's not always easy to prioritize and plan, some of those small projects are ignored for months, yet their business value are sound.
I've observed a new trend, managers who are frequently in the wait list started to use AI to generate small local apps.
They still rely on my input when it's complex, or when implementation could generate risks or need resilience and would ask for small code reviews when they are unsure of the generated code quality.
The result is win win, I have more time for high value projects the executives want to prioritize, and managers can innovate faster almost on their own.
I have started to notice some similarities to MS Access development, where an SME creates a useful app for themselves and begins to share it.
I wonder if it will have a similar pattern of creating a mess as the app starts to get uptake and the SME can't scale their attention to be an app owner, as well as an SME at the same time.
Also makes me think that an llm-developed-app-friendly shared datastore would be a useful thing to have
At my work, I am not an engineer, but work in the operations side. Our managers have a hard time getting engineering resources to do similar stuff like you were discussing.
I have been empowered to be the quick and dirty build guy for small local apps instead of fighting for engineering resources. Now our managers regularly hit me up on slack with small little items that if built, could increase productivity for their teams. I love it. I'm mostly building systems that work with google workspace (docs/sheets/forms/email). So its a lot of little appscripts, or single html file apps made available via google sites. Most of our operations staff dont have all the required underlying expertise to quickly pull this kind of thing off, and are not interested in strengthening those talents. It has allowed me to shine bright while also providing much needed relief to many teams.
How do these managers deploy the code? Is it run locally, or sent to some server?
Excel used to be, and probably still is, the primary competitor to enterprise-developed apps - a lot of businesses run on it. But, that was a locally deployed phenomenon, with an added ability to deploy it somewhere else by simply emailing the workbook to someone else.
In your organization, how do your managers turn their code into working software?
All platforms have their own known single way to deploy, they give the documentation to AI as constraints.
Some users with privileged access can run their app locally in cli or browser, else most softwares we work with can use custom modules in specific languages (html, css, js, dax, vb.net, perl, python, sql, etc.).
Ownership and trust must be established, for example only the commercial manager has access to deploy modules to CRM.
They usually are constrained to read access, unless they are informed engineering managers.
Ideally I would share pipelines to deploy static pages, or a predefined dynamic architecture. I'm wary the security risks are too great, I don't trust they would have enough time / interest to become autonomous in unconstrained environments so I didn't pursue the idea, maybe I could for static pages, or in isolated networks ..
While it's great now, my worry is that these apps will start to rot as the platforms and infra they run on advance forward, then developers will be on the hook for migrating them.
Although if they were HTML/JS/CSS with no dependencies you could argue that might be quite bullet proof since browsers have an unmatched record of backward compatibility.
With change management they can be informed of platform / infra and be autonomous to upgrade their app. They usually are in the loop if they have ownership, or will be after it happens a few times.
Also I believe the value is in the output, not the app itself. ROI of a 1 hour AI slop should™ be attained before it rots and they can just spun up a new one.
I might be hopeful and a bit selfish here, I expect my colleagues to own their toolbox.
i'm noticing this across even startups and mid-market companies too!
i don't think its going to be a silver bullet, but it doesn't need to be. niche, well understood problems with simple tooling needs are the best ones to start with.
I think the real question here isn't whether roll your own software will replace large complex 'configurable' systems, but whether companies that roll their own will replace the companies that don't.
ie are the efficiency gains of having something that's exactly tailored to you enough to create a competitive advantage.
It's back to the old idea - of software eating the world.
So for example in the UK - there is a relatively new 'energy' company called Octopus - it's grown and grown and finally overtaken the old established players.
In reality it's not an energy company - it's a software company - that used it's expertise in software to overtake it's energy supplier competitors - it was able to provide innovative products in the market because it controlled it's own software - rather than 'big vendor says no'.
I think it's telling that the founder originally left school at 16 to write computer games, before coming back to do a degree etc.
ie the question is - for any particular industry what's the benefit of custom software.
Does a bakery having it's own give it enough of an advantage?
Much of this 'configurability' is there only to allow SAAS. Normally, a mid and small business does not need to tune their IT that much all the time. This is some wet dream it IT service providers' heads. In reality most business (perhaps 80% of it) can work with very old systems, and there are so many examples that prove the statement, that I wouldn't even care to make a list.
> Does a bakery having it's own give it enough of an advantage?
50 bakeries of the same franchise may benefit from it. But it does not need to be SAP or Dynamics365 or something along this line to work. People been doing business with text-mode AS360's for ages, and nobody complained. Coca Cola was using AS-something for the warehousing in 2004, while it was already discontinued for years.
Naturally something custom creates advantage as better software mirrors better workflows. I think the more pertinent point is small companies saving money by accessing custom software on the cheap vs paying for a saas forever.
Not sure it is. Unless the Saas company is ripping you off (sure it can happen - but hopefully competition in the market would manage that over time ), then it won't be that much different from your own maintenance costs.
I always think if that's the business case for custom software ( a few quid license cost savings ) then you probably shouldn't be doing it as there is almost always a better ROI case for transformation through custom software.
So back to the bakery case. Is the benefit savings on license costs, or the fact that you can give much better estimates to customers, better de-risk supply chain issues, hire less people to operate, and improve morale via reducing busy work?
All these sort of things have to be more valuable than a few quid on licensing.
I'm a bit bitter about Octopus, although they did rescue us from the horrors of our previous supplier their insistence on us getting "smart" meters that can't then function as smart meters because of poor signal and are actually more difficult to read than our old meters has left me not hugely impressed with them.
Wasn't the smart meter rollout ( and the poor choice of hardware etc ) a central government initiative. Octopus merely used their software agility to make the most of it.
ie not sure issues with the smart meters themselves is the fault of Octopus - as the meter standards are set centrally so they can still work if you switch supplier?
Back to another old adage,
"people who are really serious about software should make their own hardware"
Octopus pressured us into getting smart meters and they're the folks I am speaking to to get the things to work so as far as I'm concerned they're on the hook.
Mind you, I couldn't help noticing that the meters themselves are owned by a leasing company... (there is a plate on each one explaining this).
I don't think it will ever really happen because of ownership.
Sure this is awesome now and maybe he shipped it in a week using AI or something, but he now owns a critical part of his wife's business. 5 years from now he is gonna be working 50/hrs a week and not want to deal with this project he barely remembers even doing, whenever an SSL cert goes bad or the CC he was paying the server bills with expires or actual bugs happen he is on the line for it.
It is lame to let family/friends pay $20/mo for something you could build in a few weeks, but they will own the product forever, I don't want to.
It doesn't have to be this messy. If I were the maker I would treat this as a good first version and transfer the ownership to the business slowly. This is just like working with any consultant.
The business is run by his wife, and if they had a SWE(-like) already, that person would’ve made this. But instead, the husband did and now owns it. He also open sourced it, so he has to live with the inevitable consequences of that too.
I wonder if a very simple moltbot can do the ongoing development on its own. I mean, the hard work is mostly out of the way. This isn't so out of the realm of capability.
That looks demonstrative! For those that don't want to click, from Aug to Feb S&P is up 10%. "Software - Applications" is down 21%.
But in this context, is Uber[9% weight, down ~4% YTD] a transportation company that roles it's own software for competitive advantage? I think other's in the composition are similar. The takeaway is maybe that the tech landscape is changing or LLMs have spooked investors and they're running without direction. But that doesn't necessarily speak to bespoke software uptake (already) cutting into profits(?) Uber would be fine in that case?
I think ERP providers should continue to exist. It's hard to vibe code financial/accounting specs and there's the IBM factor. There are a bunch of weird ways you can get things wrong.
An ERP has an established workflow that follows GAAP principles.
Hundreds of thousands of customers have cut their teeth on that workflow and improvements are metered out.
The last thing you want is to have to do PCI compliance or 1099 reporting, tax calculations for every jurisdiction. IFRS, Inventory valuation methods, SOX controls, revenue recognition rules, etc.
Not to mention if you get audited saying "Oh yeah we vibe-reconcilled all those statements".
Anything that touches the ERP? sure.
If you re-design ERPs for total AI? Maybe actual ledgers (no - not tables in MSSQL), imdepotency, rollbacks, maybe. still a bad idea.
Don't roll your own crypto. Don't roll your own ERP. Roll everything else around it.
> I think ERP providers should continue to exist. It's hard to vibe code financial/accounting specs
Perhaps you would agree that in 2026 it is fairly easy it is to actually agentically-dev a very decent ERP given one has the blueprints such as value stream diagrams, caps, BPMNs, domain models, seq. diagrams, state diagrams and... basically a complete SRS bundle. I doubt this person even needs to be super technical to deploy it.
Does it require that you use large (or small) SAAS? I guess not.
It requires one understands business architecture and know-how related to the application of such mental tools.
Traditionally it requires someone (a person) implementing these, translating them into code. Well this is precisely what LLM agentic systems do - translation. And they do it much cheaper with much shorter dev.cycles. And tailored also.
> Which as a tendency threatens ALL clumsy big ERP service providers selling you SAAS.
Wait, what?
The big ERP vendors aren't under any threat, the small ones are.
No business is going to switch from a system that has armies of low-paid consultants to in house AI developed system that has effectively zero consultants who can come in and perform the deployment with tested integrations to their accounting system, their 200 suppliers, their customer systems and their 3rd party auditing systems.
But, small businesses who were not going with a 12m contract for 5 consultants, and who dont have any need for integrations to suppliers, customers and 3rd party systems can do their own systems.
It sounds like you are very far removed from ERP and business systems in general.
All magnetic coding is going to do is further entrench existing large systems because new systems, whether AI generated or not, will be too numerous for any one of them to gain traction.
My wife's old company, a fairly significant engineering consultancy, ran it's entire time/job management and invoicing system from a company wide, custom developed Microsoft Access app called 'Time'.
It was developed by a single guy in the IT department and she liked it.
About 5 years ago the company was acquired, and they had to move to their COTS 'enterprise' system (Maconomy).
All staff from the old company had to do a week long (!) training course in how to use this and she hates it.
In future I think there will be more things like 'Time' (though presumably not MS Access based!)
> In future I think there will be more things like 'Time' (though presumably not MS Access based!)
That's my assertion - those things like 'Time' can be developed by an AI primarily because there is no requirement of an existence of a community from which to hire.
It's an example of a small ERP system - no consultants, no changes, no community, etc.
Large systems (Sage, SAP, Syspro, etc) are purchased based on the existing pool of contractors that can be hired.
Right now, if you had a competing SAP/Syspro system freshly developed, that had all the integrations that a customer needs, how on earth will they deploy it if they cannot hire people to deploy it?
Not to mention Sage's midline - Sage100 is incredibly cheap and effective for it's cost. I mean it's ridiculous what a mature software can do. Everything under the sun basically for a pittance.
It's certainly not "SAP 10 million dollar deployments". we see implementation rarely run into 6 figures for SMB distributors and manufacturing firms. That's less than most of their yearly budget for buying new fleet vehicles or equipment
I still think MS Access was awesome. In the small companies I worked it was used successfully by moderately tech savvy directors and support employees to manage ERP, license generation, invoices, etc.
The most heard gripe was the concurrent access to the database file but I think that was solved by backing the forms by accessing anything over odbc.
It looked terrible but also was highly functional.
Agreed! The first piece of software I built was a simple inventory and sales management system, around 2000. I was 16 and it was just about my first experience programming.
It was for school, and I recently found the write up and was surprised how well the system worked.
Ever since I've marvelled at how easy it was to build something highly functional that could incorporate complex business logic, and wished there was a more modern equivalent.
Grist[1] is great for this stuff, at first glance its a spreadsheet but that spreadsheet is backed by a SQLite database and you can put an actual UI on top of it without leaving the tool, or you can write full blown plugins in Javascript and HTML if you need to go further than that.
Just another yay for Grist here! I've been looking for an Access alternative for quite a while and nothing really comes close. You can try hacking it together with various BI tools, but nothing really feels as accessible as the original Access. While it's not a 1:1 mapping and the graphical report building is not really there, you can still achieve what you need. It's like Access 2.0 to me.
Access as a front end for mssqlserver ran great in a small shop. Seems like there was a wizard that imported the the access tables easily into sqlserver.
I've not seen anything as easy to use as the Access visual query builder and drag-n-drop report builder thing.
> We realized 1 million context window is not enough to explain all facets of Klarna
> Every new thread of AI is the same employee starting from scratch again. First day at the job.
Agents are limited. With you so far.
> This week we get a demo of a vibe coded frontend that is more beautiful and easy to use than any ticket management system I have seen
Again, totally matches my expectations. Agents totally make pretty stuff that looks like working software.
I just haven't drunk enough management wine to connect the dots and figure out these facts support a jira replacement.
It also gets me wondering, if Atlassian leaned more heavily into AI (More vibe coding, more agents, more layoffs) would they have been able to keep the Klarna contract?
> Are you sure about that? Because thats exactly what Klarna is doing/has done.
That link does not say that they are switching away from a system that requires armies of consultants to implement.
AFAICT, they are switching away from Jira (Atlassian/confluence products). Those are not ERP systems.
Once again, I must point out that the these sorts of assertions reveal that the person making the assertion has never been involved in an ERP rollout, neither a big one nor a small one.
And, again, I reiterate, the only threat is to small players in the market, who don't have a community to hire from. Because to become a big player, you need to gain traction as a small player, and if every small ERP system can be replaced with an AI generated system, non single one is ever going to gain traction (Why pay $10/user/month for a basic system when you can have AI generate that for a once of fee and some employee time?)
Workday is not an ERP? Beyond that, they're effectively replacing major stacks of traditional SaaS tools with in-house ones. Considering the scale and complexity of what Klarna does and the regulations it has to follow across many different markts, I'd say its a valid concern. Now, I don't think SAP etc are going anywhere, especially in traditional businesses where most of the company is reliant on it, but it seems there is a way to do it.
That said, plenty of banks still run on mainframes and use COBOL.
Well lets see how it works out for them - they're ending the partnership for HR software in order to build their own, but they say they haven't built anything yet!
> It sounds like you are very far removed from ERP and business systems in general.
am I really? it sounds so many people in big ERP service providers are oblivious of the tide rising that will wash them away, because you know what - most of these companies are super pricey, super slow, very messy and tend to fail large-scale projects that cost millions? I've seen this happen personally, and I have personally, as a sole player implemented ERPs with custom inhouse software.
Trust me brother, I know ERPs very well and seen hundreds of high-profile fakers that have zero knowledge of E/R, Business Architecture, and integrations, that still believe they can get away with nonsense.
I've been part of enough ERP contracts to know that customers evaluate their options based on how easy it is to hire consultants on the open market.
I did do quite a lot of custom software prior to AI-slop, and that market is completely destroyed with vibe-coding agents. The ERP one looks like it is simply going to further entrench the big players, because new ERPs evolve from some custom application, and once that pipeline is gone, your choices are going to be "build it yourself with no ability to hire for it" or "Go with one of the existing behemoths".
My argument is that AI will remove the pipeline that leads to incumbents seeing more competition.
> on how easy it is to hire consultants on the open market.
In the not not so distant future (5 years? 10?) that consultant market will be of people with very good process knowledge and very good prompting skills. Which might or might not be in a good part the same consultant market we have today.
My work uses these services and it’s interesting to see the divide between companies who have documented APIs (some also with “marketplaces”), and others who have many thousands of dollars partnerships license requirements to get API access.
Or companies who are very restrictive about what they will let you do with their API, presumably in an effort to control the ecosystem around their products—whether it’s because they fear being open would devalue their product, or they have a strong notion of their market position and won’t admit any integration which upsets their governance.
In another life I could have. There was an interesting professional course on healthcare systems at a local university, Hofstra. But it was all MS, and I only wanted to work with OSS.
This point on security is great point that I have not fully appreciated until now. I have been telling people that my own ability to use claude code has been a game changer for what sort of tools I will or will not pay for or use. This includes random self host services.
For personal use, those making their own software will still be a minority for a while, but at my job we are seeing potential to save $1M-$10M a year by rolling a custom tool vs paying for a commercial one. The saving here come from the tool doing a better job, not the license we pay.
I think some the prevalence of AI is actually turning the bias on this. It would actually be a return to the roots of the start of early business computing - and sort of picking up where excel et al left off. I don't think it's AI the tech itself, it's the confidence for companies to build a customized software stack and maintain it is what AI mostly contributes.
Curious how AI is giving people the confidence that they can maintain custom software. Is this Dunning-Kruger at work? It gives me great confidence in being able to stand up a new system, but as someone who has maintained software and databases for 5+ years it gives me nightmares about future maintainability.
Companies have always been able to hire software engineers as well. At least that's my impression in the UK. Is this different in other parts? Not enough engineers left after big tech has hired them all?
I've been writing custom and very specific software for a business for the past 7 years. They looked into off the shelf solutions but decided it was cheaper to build in house.
If what you predict is true then the sheer amount of software is going to explode. Good for people who actually understand how it all works!
Not just that, previously many orgs outsourced to consultancy, now when the consultancies also start outsourcing to AI, sooner all business may cut the middleman and have inhouse it teams outsourcing work to AI instead of consultancies !
You do realize that making software by developer for his wife means for random business is hiring a third party dev to build custom software?
So still, for ransom business much cheaper and better to buy software from SaaS vendor.
In this case it was better ONLY because the client is the wife of the developer.
And even now, if he sells this to other businesses - it will be MUCH cheaper to buy his subscription than homebrew the same version of it - as if it starts selling it he will be adding features and support which requires time (which is money).
> You do realize that making software by developer for his wife means for random business is hiring a third party dev to build custom software?
No, this is not true. There are so many non-technical users of Microsoft Access that run their won businesses without hiring anyone. A friend of mine had a business with an yearly turnover shy of $3M (which is small, alright) and it was running wired spreadsheets and google forms. 20 people. He never ever bought any software, and existed for more than 10 years, until his wife (yes his waifu) decided to divorce and bring the company down.
Business Architecture is not so much about writing the software, sorry, we as IT professionals would love to think it is, but this is a super weak bias.
Almost every SMB I interact with sounds like this company. Was it founded more than 10 years ago? Probably holding the ship together with spreadsheets and email.
Yep, and there are so many of these examples, as you said - most SMBs. But I know also more recent companies having it similarly.
On the other hand - we did recently pitche some brand new RAG for TBs of complex schematics for a company in the Netherlands, and guess what - they didn't like the enterprise rates that the middleman offered, not because they did not like the demo (which they loved absolutely), and not because it was late or incomplete (it took less than a month). Had I approached this company directly, with normal rates, I would be deploying it in production already. It's very telling, and not good news for large SAAS vendors.
>> was running wired spreadsheets and google forms
So he was using someone's software.
Probably he grew his business to 3M for this exact reason - by focusing on business and not playing tech company by homebrewing software. Which only proves my point.
I meant random gh repos really here, though same argument stands for SAAS which is very very very much interested in your data, and you have exactly zero guarantees and control over what happens to it once ingested. Like - GDPR exists not to protect users, but to enforce anonymization of the data that is then going to be aggregated and sold.
Now imagine my not-so-complext ERP or internal system - can be developed with little or no effort. Why would I give Benioff my dollars rather than spend it on in-house assets, that also increase the valuation of my company? I find very little reason to do so in 2026.
This is a great piece of software, with much thought put into nitty gritty details. Aside from the gripes around the mobile experience that some have outlined here, I would say you've put much thought into this piece of software. Your wife is lucky that she has a talented software developer for a husband. AI or no AI, I think this is a very clean and beautiful piece of software. This doesn't seem like its Vibe coded, because AI doesn't write such clean code but maybe AI is improving and I'm just bad at telling which is which. Nonetheless, keep up the great work and thanks for sharing. I'm downloading it just to learn from your codebase. Its not like before AI came around talented devs didn't create working side projects to help their loved ones out.
The truth is that I put a lot of work in at the beginning to define the data structure and flows by myself. AI was very useful later to experiment with how to build the views on top of it and to fix some issues.
I still had to think hard about how to make this simple and easy for someone who does not have deep knowledge of this manufacturing domain.
*Lessons learned:*
1. Data structure is almost everything, then comes business logic
2. You must have deep domain knowledge of what you are building
3. Iterate fast on the views built on top of the data structure
*PS*
All mobile fixed had been resolved and deployed ;)
"Show me your flowchart and conceal your tables, and I shall continue to be mystified. Show me your tables, and I won't usually need your flowchart; it'll be obvious." -- Fred Brooks
AI will write whatever kind of code you want it to write. If you know how to write clean code and you can describe that in a prompt, it will give you clean code that is indistinguishable from that of a talented dev.
I find it less time consuming to just write the code, understand it 99% (since I wrote it), and debug the rest, than it is to try to describe it to the AI, understand a fraction of what it spits out, and spend more time understanding it and fixing it.
If you can just write clean code just do that. Also, you will improve your skills even more the more you do that (shocker). So the next time you have to do that it will be even easier. This is called learning a skill.
Sorry for the rough tone, reading that back haha. But still posting because I'm just passionate about it, it's nothing personal though.
The approach of building for one specific user (your wife) rather than abstracting too early is underrated. You end up with something that actually fits the workflows instead of a generic tool that needs heavy configuration.
Elixir + Ash is an interesting choice for this domain. LiveView particularly shines for internal tools like this where you want the interactivity without managing a separate frontend build. Curious how the AI code generation worked with Ash specifically - the declarative nature seems like it could either help a lot (clear patterns) or confuse models that expect more explicit code.
The BOM with cost rollups is the feature that would have saved me hours in a previous project. Most small batch producers I know either overprice everything out of caution or underestimate costs because tracking ingredient pricing through recipes is tedious in spreadsheets.
I love this so much. I do worry about the hidden cost of maintaining these software stacks that have a userbase of one. Things like hidden vulnerabilities, bugs, and edge cases that could bring a business down for days.
But I do think the days of just having to learn, tolerate, and accept any commercial software stack for your business because it's too complex to build yourself are over. What vendors remain will have to absolutely meet users with their unique requirements and budget.
Elixir + Ash is so good, and AI works very well with it because it can understand the domain you are working in. LiveView is amazing, not just for internal tools. I would even say that in most cases it will outperform flashy UIs, because it focuses on actual users who work with the app every day, not people who just want to be impressed for a minute
I relate to this alot. I've been working in B2B SaaS for years and saw the exact same gap in customer success tooling. The serious tools start at $10k - $50k per year and take months to set up, and everyone else that can't afford them use spreadsheets. There's almost nothing in the middle.
What's funny is the actual workflows usually aren't that complicated. The complexity is artificial. It comes from the vendor trying to serve every possible use case. When you build something focused on the real workflow, it turns out you can get 80% of the value in a fraction of the time and cost.
The open source angle is smart tool. In B2B, the "let me just try it" factor matter so much. If someone has to sit through a sales call just to see if the tool fits, most small operators will just stick with their spreadsheet.
I love this. I know another small batch baker who also thought it was cool (we'll dig more when sober). BOM+cost is rad. Eager to try forecasting a weekend-rush situation.
My only nit, as a legacy internet goober is, use example.com for these throw away addresses; it's reserved for that purpose.
This is amazing! My wife is also planning to open up a bakery and I was thinking of building something similar. Hadn’t thought of that daily production workflow, just the ingredients and recipes into cost/labour/profit parts.
Some small things:
When trying to edit a product (almond cookies) on the phone, I cannot scroll the pop-up so cannot go to all fields or the save button.
When calculating the total calories it prints kg as the unit instead of cal.
On the overview page of materials for a product it only shows “grams” per ingredient without the actual number.
Is the logic behind "Usage Forecast" and "Reorder Planner" hard-coded somewhere? I'm not seeing any configuration for that, so I had to ask the question.
Congrats, deofoo. Very interesting. How much of it was developed with AI? If you had to give us a percentage of how much of the code was written by yourself vs AI, what would the parity be?
Awesome, does it support Semiproducts? For example I'm making my own jam (sugar + fruit) and then I use this again as a material in different finished products, say cakes and croissants.
You can tell this was written with love. Regardless of what tools were used to make this, I think we could all do more of this for the people important in our lives.
Super cool. My wife and me recently analyzed a small bagel shop that was open for sale in Berlin. Unfortunately it didn’t work out for us.
But such a software would have been incredibly useful. Former owner did everything in .xls.
I really like this UI. Just ordered an Elixir book. I'm excited to learn this stack in the coming weeks. Hopefully it'll encourage me to build an app I really want for personal use. I can't bring myself to use the stack I use for work at home.
I think this is a very nicely thought out approach. I particularly like it doing allergen tracking. Obviously you're at the mercy of supplier/supply-chain integrity but if you do e.g. wind up with ground cumin contaminated with god knows what, this is what will get you where you need to be.
Just amazing. I had a need for something like this but wound up building it out in Mathesar. That, too, is an amazing project, but my business logic has to remain separate. Jeepers, you're even getting into labor accounting - well done!
Love to the point of invention! This looks and feels great.
I'm an Elixir newbie and wondering if I should start with learning Ash or stick with Liveview until I know more. Any thoughts on what Ash solved for you over Phoenix Liveview?
Ash can be used in conjunction with Phoenix, they aren't mutually exclusive. Ash is really just a framework for modeling your domain(s) and getting a bunch of helpful functionality for free (e.g advanced querying capabilities, pagination, data validations, json+graphql apis, and more) Then you could use those functionalities with phoenix to build a full web app. Or you could use something else other than phoenix, it's up to you :)
I'm actually building open-source SaaS for every vertical and leveraging that to build an interoperable marketplace. I'm nearly done with the restaurant one (alternative to Toast).
This looks great! Love the screenshots, although I'm unable to login to the app. Likely due to the extra traffic this post has gotten.
How long did it take you to build this ?
First time hearing the term micro bakery. Is there an equivalent for restaurants? Like a micro restaurant? Seems like a fun idea for my retired parents.
Looks well thought out. We wrestle with website, real ERP and building Notion connectors for production orders in make to order scenarios so there’s definitely a pain point.
Honestly, well done and thanks for sharing it. I also really appreciate the fact that you included multiple screenshots of the UI, as well as some of the agent plans. Reading the code and project structure, it feels like you put in the work.
this is sick, my wife complains about her franchise's mandatory technology portal bs etc. I was thinking I could definitely whip up something better and you've inspired me more.
Good, I will with great pleasure now reiterate my point about people now producing their own code, even complex stuff, rather than downloading potentially malicious and foreign code. Which as a tendency threatens ALL clumsy big ERP service providers selling you SAAS.
Go ahead - I'm ready to be down-voted again and again until folks realize it is inevitable, as is inevitable that many companies in the area of business software are going down down down.
I'll down vote just for the whiny attitude and superior tone.
I'm the main dev in a small IT company, my backlog is filled with requests, it's not always easy to prioritize and plan, some of those small projects are ignored for months, yet their business value are sound.
I've observed a new trend, managers who are frequently in the wait list started to use AI to generate small local apps. They still rely on my input when it's complex, or when implementation could generate risks or need resilience and would ask for small code reviews when they are unsure of the generated code quality.
The result is win win, I have more time for high value projects the executives want to prioritize, and managers can innovate faster almost on their own.
I have started to notice some similarities to MS Access development, where an SME creates a useful app for themselves and begins to share it.
I wonder if it will have a similar pattern of creating a mess as the app starts to get uptake and the SME can't scale their attention to be an app owner, as well as an SME at the same time.
Also makes me think that an llm-developed-app-friendly shared datastore would be a useful thing to have
I like your 'llm-developped-app-friendly shared datastore' idea and would pursue it.
While taking ownership of AI slop is not an option for me, I do want to avoid shadow IT.
Might be a moon shot .. could sharing a prompt template with git access to colleagues be a way to enforce it ?
At my work, I am not an engineer, but work in the operations side. Our managers have a hard time getting engineering resources to do similar stuff like you were discussing.
I have been empowered to be the quick and dirty build guy for small local apps instead of fighting for engineering resources. Now our managers regularly hit me up on slack with small little items that if built, could increase productivity for their teams. I love it. I'm mostly building systems that work with google workspace (docs/sheets/forms/email). So its a lot of little appscripts, or single html file apps made available via google sites. Most of our operations staff dont have all the required underlying expertise to quickly pull this kind of thing off, and are not interested in strengthening those talents. It has allowed me to shine bright while also providing much needed relief to many teams.
How do these managers deploy the code? Is it run locally, or sent to some server?
Excel used to be, and probably still is, the primary competitor to enterprise-developed apps - a lot of businesses run on it. But, that was a locally deployed phenomenon, with an added ability to deploy it somewhere else by simply emailing the workbook to someone else.
In your organization, how do your managers turn their code into working software?
All platforms have their own known single way to deploy, they give the documentation to AI as constraints.
Some users with privileged access can run their app locally in cli or browser, else most softwares we work with can use custom modules in specific languages (html, css, js, dax, vb.net, perl, python, sql, etc.). Ownership and trust must be established, for example only the commercial manager has access to deploy modules to CRM. They usually are constrained to read access, unless they are informed engineering managers.
Ideally I would share pipelines to deploy static pages, or a predefined dynamic architecture. I'm wary the security risks are too great, I don't trust they would have enough time / interest to become autonomous in unconstrained environments so I didn't pursue the idea, maybe I could for static pages, or in isolated networks ..
Thanks for the reply. Appreciated.
While it's great now, my worry is that these apps will start to rot as the platforms and infra they run on advance forward, then developers will be on the hook for migrating them.
Although if they were HTML/JS/CSS with no dependencies you could argue that might be quite bullet proof since browsers have an unmatched record of backward compatibility.
With change management they can be informed of platform / infra and be autonomous to upgrade their app. They usually are in the loop if they have ownership, or will be after it happens a few times.
Also I believe the value is in the output, not the app itself. ROI of a 1 hour AI slop should™ be attained before it rots and they can just spun up a new one.
I might be hopeful and a bit selfish here, I expect my colleagues to own their toolbox.
i'm noticing this across even startups and mid-market companies too!
i don't think its going to be a silver bullet, but it doesn't need to be. niche, well understood problems with simple tooling needs are the best ones to start with.
https://culturecompiled.com/p/things-are-getting-awkward-for...
That sounds good enough for me.
I think the real question here isn't whether roll your own software will replace large complex 'configurable' systems, but whether companies that roll their own will replace the companies that don't.
ie are the efficiency gains of having something that's exactly tailored to you enough to create a competitive advantage.
It's back to the old idea - of software eating the world.
So for example in the UK - there is a relatively new 'energy' company called Octopus - it's grown and grown and finally overtaken the old established players.
In reality it's not an energy company - it's a software company - that used it's expertise in software to overtake it's energy supplier competitors - it was able to provide innovative products in the market because it controlled it's own software - rather than 'big vendor says no'.
I think it's telling that the founder originally left school at 16 to write computer games, before coming back to do a degree etc.
ie the question is - for any particular industry what's the benefit of custom software. Does a bakery having it's own give it enough of an advantage?
Much of this 'configurability' is there only to allow SAAS. Normally, a mid and small business does not need to tune their IT that much all the time. This is some wet dream it IT service providers' heads. In reality most business (perhaps 80% of it) can work with very old systems, and there are so many examples that prove the statement, that I wouldn't even care to make a list.
> Does a bakery having it's own give it enough of an advantage? 50 bakeries of the same franchise may benefit from it. But it does not need to be SAP or Dynamics365 or something along this line to work. People been doing business with text-mode AS360's for ages, and nobody complained. Coca Cola was using AS-something for the warehousing in 2004, while it was already discontinued for years.
"Much of this 'configurability' is there only to allow SAAS"
ERPs and CRMs were highly configurable long before SaaS.
Naturally something custom creates advantage as better software mirrors better workflows. I think the more pertinent point is small companies saving money by accessing custom software on the cheap vs paying for a saas forever.
> I think the more pertinent point
Not sure it is. Unless the Saas company is ripping you off (sure it can happen - but hopefully competition in the market would manage that over time ), then it won't be that much different from your own maintenance costs.
I always think if that's the business case for custom software ( a few quid license cost savings ) then you probably shouldn't be doing it as there is almost always a better ROI case for transformation through custom software.
So back to the bakery case. Is the benefit savings on license costs, or the fact that you can give much better estimates to customers, better de-risk supply chain issues, hire less people to operate, and improve morale via reducing busy work?
All these sort of things have to be more valuable than a few quid on licensing.
however, many workflows are part of generally accepted accounting principles, or GAAP for short.
Many workflows are about B2B transactions. PO -> Sales Order -> Invoicing workflows. ASNs, etc.
So a lot of workflows are not driven by companies but by the standard operating framework of B2B
I'm a bit bitter about Octopus, although they did rescue us from the horrors of our previous supplier their insistence on us getting "smart" meters that can't then function as smart meters because of poor signal and are actually more difficult to read than our old meters has left me not hugely impressed with them.
Wasn't the smart meter rollout ( and the poor choice of hardware etc ) a central government initiative. Octopus merely used their software agility to make the most of it.
ie not sure issues with the smart meters themselves is the fault of Octopus - as the meter standards are set centrally so they can still work if you switch supplier?
Back to another old adage,
"people who are really serious about software should make their own hardware"
Octopus pressured us into getting smart meters and they're the folks I am speaking to to get the things to work so as far as I'm concerned they're on the hook.
Mind you, I couldn't help noticing that the meters themselves are owned by a leasing company... (there is a plate on each one explaining this).
Again I think that's mostly the governments fault - they are setting mandatory smart meter rollout targets for the energy suppliers every year.
I don't think it will ever really happen because of ownership.
Sure this is awesome now and maybe he shipped it in a week using AI or something, but he now owns a critical part of his wife's business. 5 years from now he is gonna be working 50/hrs a week and not want to deal with this project he barely remembers even doing, whenever an SSL cert goes bad or the CC he was paying the server bills with expires or actual bugs happen he is on the line for it.
It is lame to let family/friends pay $20/mo for something you could build in a few weeks, but they will own the product forever, I don't want to.
Many times we're already on the hook anyways, supporting friends/family even when they are using someone else's product.
It doesn't have to be this messy. If I were the maker I would treat this as a good first version and transfer the ownership to the business slowly. This is just like working with any consultant.
The business is run by his wife, and if they had a SWE(-like) already, that person would’ve made this. But instead, the husband did and now owns it. He also open sourced it, so he has to live with the inevitable consequences of that too.
I wonder if a very simple moltbot can do the ongoing development on its own. I mean, the hard work is mostly out of the way. This isn't so out of the realm of capability.
in 5 years whatever ai tools will be good enough to have ownership of a critical piece of software that was built now by ai.
First slowly then all at once. [1]
[1]: https://finance.yahoo.com/sectors/technology/software-applic...
That looks demonstrative! For those that don't want to click, from Aug to Feb S&P is up 10%. "Software - Applications" is down 21%.
But in this context, is Uber[9% weight, down ~4% YTD] a transportation company that roles it's own software for competitive advantage? I think other's in the composition are similar. The takeaway is maybe that the tech landscape is changing or LLMs have spooked investors and they're running without direction. But that doesn't necessarily speak to bespoke software uptake (already) cutting into profits(?) Uber would be fine in that case?
I think ERP providers should continue to exist. It's hard to vibe code financial/accounting specs and there's the IBM factor. There are a bunch of weird ways you can get things wrong.
An ERP has an established workflow that follows GAAP principles.
Hundreds of thousands of customers have cut their teeth on that workflow and improvements are metered out.
The last thing you want is to have to do PCI compliance or 1099 reporting, tax calculations for every jurisdiction. IFRS, Inventory valuation methods, SOX controls, revenue recognition rules, etc.
Not to mention if you get audited saying "Oh yeah we vibe-reconcilled all those statements".
Anything that touches the ERP? sure.
If you re-design ERPs for total AI? Maybe actual ledgers (no - not tables in MSSQL), imdepotency, rollbacks, maybe. still a bad idea.
Don't roll your own crypto. Don't roll your own ERP. Roll everything else around it.
> I think ERP providers should continue to exist. It's hard to vibe code financial/accounting specs
Perhaps you would agree that in 2026 it is fairly easy it is to actually agentically-dev a very decent ERP given one has the blueprints such as value stream diagrams, caps, BPMNs, domain models, seq. diagrams, state diagrams and... basically a complete SRS bundle. I doubt this person even needs to be super technical to deploy it.
Does it require that you use large (or small) SAAS? I guess not.
It requires one understands business architecture and know-how related to the application of such mental tools.
Traditionally it requires someone (a person) implementing these, translating them into code. Well this is precisely what LLM agentic systems do - translation. And they do it much cheaper with much shorter dev.cycles. And tailored also.
I don't think you need a SaaS. I think you need an ERP.
Point made was about SaaS companies selling ERP in the general sense. Otherwise I agree - nosy of the times SMBs don’t need SaaS.
> Which as a tendency threatens ALL clumsy big ERP service providers selling you SAAS.
Wait, what?
The big ERP vendors aren't under any threat, the small ones are.
No business is going to switch from a system that has armies of low-paid consultants to in house AI developed system that has effectively zero consultants who can come in and perform the deployment with tested integrations to their accounting system, their 200 suppliers, their customer systems and their 3rd party auditing systems.
But, small businesses who were not going with a 12m contract for 5 consultants, and who dont have any need for integrations to suppliers, customers and 3rd party systems can do their own systems.
It sounds like you are very far removed from ERP and business systems in general.
All magnetic coding is going to do is further entrench existing large systems because new systems, whether AI generated or not, will be too numerous for any one of them to gain traction.
My wife's old company, a fairly significant engineering consultancy, ran it's entire time/job management and invoicing system from a company wide, custom developed Microsoft Access app called 'Time'.
It was developed by a single guy in the IT department and she liked it.
About 5 years ago the company was acquired, and they had to move to their COTS 'enterprise' system (Maconomy).
All staff from the old company had to do a week long (!) training course in how to use this and she hates it.
In future I think there will be more things like 'Time' (though presumably not MS Access based!)
> In future I think there will be more things like 'Time' (though presumably not MS Access based!)
That's my assertion - those things like 'Time' can be developed by an AI primarily because there is no requirement of an existence of a community from which to hire.
It's an example of a small ERP system - no consultants, no changes, no community, etc.
Large systems (Sage, SAP, Syspro, etc) are purchased based on the existing pool of contractors that can be hired.
Right now, if you had a competing SAP/Syspro system freshly developed, that had all the integrations that a customer needs, how on earth will they deploy it if they cannot hire people to deploy it?
Not to mention Sage's midline - Sage100 is incredibly cheap and effective for it's cost. I mean it's ridiculous what a mature software can do. Everything under the sun basically for a pittance.
It's certainly not "SAP 10 million dollar deployments". we see implementation rarely run into 6 figures for SMB distributors and manufacturing firms. That's less than most of their yearly budget for buying new fleet vehicles or equipment
I still think MS Access was awesome. In the small companies I worked it was used successfully by moderately tech savvy directors and support employees to manage ERP, license generation, invoices, etc.
The most heard gripe was the concurrent access to the database file but I think that was solved by backing the forms by accessing anything over odbc.
It looked terrible but also was highly functional.
Agreed! The first piece of software I built was a simple inventory and sales management system, around 2000. I was 16 and it was just about my first experience programming.
It was for school, and I recently found the write up and was surprised how well the system worked.
Ever since I've marvelled at how easy it was to build something highly functional that could incorporate complex business logic, and wished there was a more modern equivalent.
Maybe a combination of AirTable and PowerBI/open-source alternative? Or just ms access backed by a proper database?
Grist[1] is great for this stuff, at first glance its a spreadsheet but that spreadsheet is backed by a SQLite database and you can put an actual UI on top of it without leaving the tool, or you can write full blown plugins in Javascript and HTML if you need to go further than that.
[1] https://www.getgrist.com/
Just another yay for Grist here! I've been looking for an Access alternative for quite a while and nothing really comes close. You can try hacking it together with various BI tools, but nothing really feels as accessible as the original Access. While it's not a 1:1 mapping and the graphical report building is not really there, you can still achieve what you need. It's like Access 2.0 to me.
Access as a front end for mssqlserver ran great in a small shop. Seems like there was a wizard that imported the the access tables easily into sqlserver.
I've not seen anything as easy to use as the Access visual query builder and drag-n-drop report builder thing.
> No business is going to switch from a system that has armies of low-paid consultants to in house AI developed system
Are you sure about that? Because thats exactly what Klarna is doing/has done.
https://threadreaderapp.com/thread/1957789124930286065.html?...
> We realized 1 million context window is not enough to explain all facets of Klarna > Every new thread of AI is the same employee starting from scratch again. First day at the job.
Agents are limited. With you so far.
> This week we get a demo of a vibe coded frontend that is more beautiful and easy to use than any ticket management system I have seen
Again, totally matches my expectations. Agents totally make pretty stuff that looks like working software.
I just haven't drunk enough management wine to connect the dots and figure out these facts support a jira replacement.
It also gets me wondering, if Atlassian leaned more heavily into AI (More vibe coding, more agents, more layoffs) would they have been able to keep the Klarna contract?
> Are you sure about that? Because thats exactly what Klarna is doing/has done.
That link does not say that they are switching away from a system that requires armies of consultants to implement.
AFAICT, they are switching away from Jira (Atlassian/confluence products). Those are not ERP systems.
Once again, I must point out that the these sorts of assertions reveal that the person making the assertion has never been involved in an ERP rollout, neither a big one nor a small one.
And, again, I reiterate, the only threat is to small players in the market, who don't have a community to hire from. Because to become a big player, you need to gain traction as a small player, and if every small ERP system can be replaced with an AI generated system, non single one is ever going to gain traction (Why pay $10/user/month for a basic system when you can have AI generate that for a once of fee and some employee time?)
Workday is not an ERP? Beyond that, they're effectively replacing major stacks of traditional SaaS tools with in-house ones. Considering the scale and complexity of what Klarna does and the regulations it has to follow across many different markts, I'd say its a valid concern. Now, I don't think SAP etc are going anywhere, especially in traditional businesses where most of the company is reliant on it, but it seems there is a way to do it.
That said, plenty of banks still run on mainframes and use COBOL.
https://www.salesforceben.com/klarna-salesforce-workday-part...
Well lets see how it works out for them - they're ending the partnership for HR software in order to build their own, but they say they haven't built anything yet!
> It sounds like you are very far removed from ERP and business systems in general.
am I really? it sounds so many people in big ERP service providers are oblivious of the tide rising that will wash them away, because you know what - most of these companies are super pricey, super slow, very messy and tend to fail large-scale projects that cost millions? I've seen this happen personally, and I have personally, as a sole player implemented ERPs with custom inhouse software.
Trust me brother, I know ERPs very well and seen hundreds of high-profile fakers that have zero knowledge of E/R, Business Architecture, and integrations, that still believe they can get away with nonsense.
Hope you're not one of them.
> Hope you're not one of them.
I've been part of enough ERP contracts to know that customers evaluate their options based on how easy it is to hire consultants on the open market.
I did do quite a lot of custom software prior to AI-slop, and that market is completely destroyed with vibe-coding agents. The ERP one looks like it is simply going to further entrench the big players, because new ERPs evolve from some custom application, and once that pipeline is gone, your choices are going to be "build it yourself with no ability to hire for it" or "Go with one of the existing behemoths".
My argument is that AI will remove the pipeline that leads to incumbents seeing more competition.
> on how easy it is to hire consultants on the open market.
In the not not so distant future (5 years? 10?) that consultant market will be of people with very good process knowledge and very good prompting skills. Which might or might not be in a good part the same consultant market we have today.
You think the big players aren’t going to use the new tools to massively cut employees and costs?
> “…big ERP service providers selling you SAAS.”
My work uses these services and it’s interesting to see the divide between companies who have documented APIs (some also with “marketplaces”), and others who have many thousands of dollars partnerships license requirements to get API access.
Or companies who are very restrictive about what they will let you do with their API, presumably in an effort to control the ecosystem around their products—whether it’s because they fear being open would devalue their product, or they have a strong notion of their market position and won’t admit any integration which upsets their governance.
Do you work in the healthcare space? Sounds like my issues
In another life I could have. There was an interesting professional course on healthcare systems at a local university, Hofstra. But it was all MS, and I only wanted to work with OSS.
No. I’m in financial sector, accounting.
This point on security is great point that I have not fully appreciated until now. I have been telling people that my own ability to use claude code has been a game changer for what sort of tools I will or will not pay for or use. This includes random self host services.
For personal use, those making their own software will still be a minority for a while, but at my job we are seeing potential to save $1M-$10M a year by rolling a custom tool vs paying for a commercial one. The saving here come from the tool doing a better job, not the license we pay.
I think some the prevalence of AI is actually turning the bias on this. It would actually be a return to the roots of the start of early business computing - and sort of picking up where excel et al left off. I don't think it's AI the tech itself, it's the confidence for companies to build a customized software stack and maintain it is what AI mostly contributes.
Curious how AI is giving people the confidence that they can maintain custom software. Is this Dunning-Kruger at work? It gives me great confidence in being able to stand up a new system, but as someone who has maintained software and databases for 5+ years it gives me nightmares about future maintainability.
Companies have always been able to hire software engineers as well. At least that's my impression in the UK. Is this different in other parts? Not enough engineers left after big tech has hired them all?
I've been writing custom and very specific software for a business for the past 7 years. They looked into off the shelf solutions but decided it was cheaper to build in house.
If what you predict is true then the sheer amount of software is going to explode. Good for people who actually understand how it all works!
Not just that, previously many orgs outsourced to consultancy, now when the consultancies also start outsourcing to AI, sooner all business may cut the middleman and have inhouse it teams outsourcing work to AI instead of consultancies !
You do realize that making software by developer for his wife means for random business is hiring a third party dev to build custom software?
So still, for ransom business much cheaper and better to buy software from SaaS vendor.
In this case it was better ONLY because the client is the wife of the developer.
And even now, if he sells this to other businesses - it will be MUCH cheaper to buy his subscription than homebrew the same version of it - as if it starts selling it he will be adding features and support which requires time (which is money).
> You do realize that making software by developer for his wife means for random business is hiring a third party dev to build custom software?
No, this is not true. There are so many non-technical users of Microsoft Access that run their won businesses without hiring anyone. A friend of mine had a business with an yearly turnover shy of $3M (which is small, alright) and it was running wired spreadsheets and google forms. 20 people. He never ever bought any software, and existed for more than 10 years, until his wife (yes his waifu) decided to divorce and bring the company down.
Business Architecture is not so much about writing the software, sorry, we as IT professionals would love to think it is, but this is a super weak bias.
Almost every SMB I interact with sounds like this company. Was it founded more than 10 years ago? Probably holding the ship together with spreadsheets and email.
Yep, and there are so many of these examples, as you said - most SMBs. But I know also more recent companies having it similarly.
On the other hand - we did recently pitche some brand new RAG for TBs of complex schematics for a company in the Netherlands, and guess what - they didn't like the enterprise rates that the middleman offered, not because they did not like the demo (which they loved absolutely), and not because it was late or incomplete (it took less than a month). Had I approached this company directly, with normal rates, I would be deploying it in production already. It's very telling, and not good news for large SAAS vendors.
>> He never ever bought any software
>> was running wired spreadsheets and google forms
So he was using someone's software.
Probably he grew his business to 3M for this exact reason - by focusing on business and not playing tech company by homebrewing software. Which only proves my point.
So true, we devs are so stuck in our own bubble
> rather than downloading potentially malicious and foreign code
So I shouldn't download and use this right? I can't verify if it's potentially malicious or not
I meant random gh repos really here, though same argument stands for SAAS which is very very very much interested in your data, and you have exactly zero guarantees and control over what happens to it once ingested. Like - GDPR exists not to protect users, but to enforce anonymization of the data that is then going to be aggregated and sold.
Now imagine my not-so-complext ERP or internal system - can be developed with little or no effort. Why would I give Benioff my dollars rather than spend it on in-house assets, that also increase the valuation of my company? I find very little reason to do so in 2026.
100%
As if the AI generated slop isn't downloading potentially malicious and foreign code.
Dude. Stay on topic.
This is a great piece of software, with much thought put into nitty gritty details. Aside from the gripes around the mobile experience that some have outlined here, I would say you've put much thought into this piece of software. Your wife is lucky that she has a talented software developer for a husband. AI or no AI, I think this is a very clean and beautiful piece of software. This doesn't seem like its Vibe coded, because AI doesn't write such clean code but maybe AI is improving and I'm just bad at telling which is which. Nonetheless, keep up the great work and thanks for sharing. I'm downloading it just to learn from your codebase. Its not like before AI came around talented devs didn't create working side projects to help their loved ones out.
The truth is that I put a lot of work in at the beginning to define the data structure and flows by myself. AI was very useful later to experiment with how to build the views on top of it and to fix some issues.
I still had to think hard about how to make this simple and easy for someone who does not have deep knowledge of this manufacturing domain.
*Lessons learned:*
*PS*All mobile fixed had been resolved and deployed ;)
#1 reminds me of my favourite software quote:
"Show me your flowchart and conceal your tables, and I shall continue to be mystified. Show me your tables, and I won't usually need your flowchart; it'll be obvious." -- Fred Brooks
AI will write whatever kind of code you want it to write. If you know how to write clean code and you can describe that in a prompt, it will give you clean code that is indistinguishable from that of a talented dev.
> If you know how to write clean code.
If we're assuming that:
I find it less time consuming to just write the code, understand it 99% (since I wrote it), and debug the rest, than it is to try to describe it to the AI, understand a fraction of what it spits out, and spend more time understanding it and fixing it.
If you can just write clean code just do that. Also, you will improve your skills even more the more you do that (shocker). So the next time you have to do that it will be even easier. This is called learning a skill.
Sorry for the rough tone, reading that back haha. But still posting because I'm just passionate about it, it's nothing personal though.
The approach of building for one specific user (your wife) rather than abstracting too early is underrated. You end up with something that actually fits the workflows instead of a generic tool that needs heavy configuration.
Elixir + Ash is an interesting choice for this domain. LiveView particularly shines for internal tools like this where you want the interactivity without managing a separate frontend build. Curious how the AI code generation worked with Ash specifically - the declarative nature seems like it could either help a lot (clear patterns) or confuse models that expect more explicit code.
The BOM with cost rollups is the feature that would have saved me hours in a previous project. Most small batch producers I know either overprice everything out of caution or underestimate costs because tracking ingredient pricing through recipes is tedious in spreadsheets.
I love this so much. I do worry about the hidden cost of maintaining these software stacks that have a userbase of one. Things like hidden vulnerabilities, bugs, and edge cases that could bring a business down for days.
But I do think the days of just having to learn, tolerate, and accept any commercial software stack for your business because it's too complex to build yourself are over. What vendors remain will have to absolutely meet users with their unique requirements and budget.
Elixir + Ash is so good, and AI works very well with it because it can understand the domain you are working in. LiveView is amazing, not just for internal tools. I would even say that in most cases it will outperform flashy UIs, because it focuses on actual users who work with the app every day, not people who just want to be impressed for a minute
OK HN, time for us to build a full open source general purpose ERP in Elixir based on Ash XD
"That's a great idea! Do you want me to create a PRD, a million markdown files and some useless Playwright tests?" -- Claude
... and you don't have token credits anymore.
AI Agents Assemble!
I relate to this alot. I've been working in B2B SaaS for years and saw the exact same gap in customer success tooling. The serious tools start at $10k - $50k per year and take months to set up, and everyone else that can't afford them use spreadsheets. There's almost nothing in the middle. What's funny is the actual workflows usually aren't that complicated. The complexity is artificial. It comes from the vendor trying to serve every possible use case. When you build something focused on the real workflow, it turns out you can get 80% of the value in a fraction of the time and cost. The open source angle is smart tool. In B2B, the "let me just try it" factor matter so much. If someone has to sit through a sales call just to see if the tool fits, most small operators will just stick with their spreadsheet.
And I built e-commerce for my wife's micro-bakery https://thonon-les-pains.fr/ (most of it - like product and order management - is behind auth).
I don't think it's useful to anyone - not white label, not open source - but still funny :)
I don’t live nearby, but thank her for baking gluten free!
The best part is the domain, I grew up close by
(Le filtre pain/biscuits/gluten est pété sous safari)
Yep the wordplay made me chuckle too :)
Génial comme nom :)
Love that. I was thinking to add e-commerce layer
I love this. I know another small batch baker who also thought it was cool (we'll dig more when sober). BOM+cost is rad. Eager to try forecasting a weekend-rush situation.
My only nit, as a legacy internet goober is, use example.com for these throw away addresses; it's reserved for that purpose.
Thanks! Would love to hear your feedback. I'm a solo developer so iteration is really fast
You could have called it craftpan (pan = bread in spanish :D)
This is amazing! My wife is also planning to open up a bakery and I was thinking of building something similar. Hadn’t thought of that daily production workflow, just the ingredients and recipes into cost/labour/profit parts.
Some small things: When trying to edit a product (almond cookies) on the phone, I cannot scroll the pop-up so cannot go to all fields or the save button. When calculating the total calories it prints kg as the unit instead of cal. On the overview page of materials for a product it only shows “grams” per ingredient without the actual number.
Super! Will get these fixed asap
Fixed!
Oh man. My wife's biscotti business will benefit from this. Nice work!
Awesome. Please send any feedback, it will make me so happy to get another person to benefit from it
This is definitely a nit but is there any reason you need 2 decimal places accuracy for percent complete?
Nope, might change it based on the feedback
Is the logic behind "Usage Forecast" and "Reorder Planner" hard-coded somewhere? I'm not seeing any configuration for that, so I had to ask the question.
Right now it's hard-coded. But based on your feedback I'm going to make some of it configurable
I think I have needed this for 3d printing for some time
The question is, does the bakery still produce bread, or does it now randomly produce bread, ice cream or frogs?
I’d consider the latter scenario a feature :)
This is very cool! How did you make a dashboard this good?
Congrats, deofoo. Very interesting. How much of it was developed with AI? If you had to give us a percentage of how much of the code was written by yourself vs AI, what would the parity be?
Awesome, does it support Semiproducts? For example I'm making my own jam (sugar + fruit) and then I use this again as a material in different finished products, say cakes and croissants.
Nice idea, i think i can make it work. Give me a day or two
Can you estimate how many hours went into this? Did you use agents?
A few weekends. At the end I used AI to get JSON:API and GraphQL working, plus the docs website, a few UI changes + bug fixing
Just doing it in Emacs is also an option.
https://news.ycombinator.com/item?id=19252952
You can tell this was written with love. Regardless of what tools were used to make this, I think we could all do more of this for the people important in our lives.
It can still be a LLM-induced marketing stunt to catch attention though. "Hey let's add a bakery story so people can relate to it!"
Assuming this was user tested in hand with your wife, it’s nice to this kind of project built for a real user.
So often, these tools lack usability because they’re built generically any use case. Here, it was designed for your wife’s bakery.
The process to build a site/app like this will only get easier and more defect-free over the coming months.
Nice work. My wife will love this. Only minor gripe is the production schedule doesn't play nicely on mobile. Text leaks between columns.
Oh man, going to fix that ASAP
Fixed!
Super cool. My wife and me recently analyzed a small bagel shop that was open for sale in Berlin. Unfortunately it didn’t work out for us. But such a software would have been incredibly useful. Former owner did everything in .xls.
I really like this UI. Just ordered an Elixir book. I'm excited to learn this stack in the coming weeks. Hopefully it'll encourage me to build an app I really want for personal use. I can't bring myself to use the stack I use for work at home.
Looks like it shares many characteristics with LIMS (Laboratory Information Management System)...
omg its built with elixir... Everytime I see my favorite language pop up I'm a little happier.
I love that you can just "curl download the docker compose file and docker up" to run. Awesome UX.
I think this is a very nicely thought out approach. I particularly like it doing allergen tracking. Obviously you're at the mercy of supplier/supply-chain integrity but if you do e.g. wind up with ground cumin contaminated with god knows what, this is what will get you where you need to be.
Yeah. Craftplan support batches so you can easily track down a bad batch
Nice! 5 bucks says you can swap this in for your average software kanban and it does a better job.
Just amazing. I had a need for something like this but wound up building it out in Mathesar. That, too, is an amazing project, but my business logic has to remain separate. Jeepers, you're even getting into labor accounting - well done!
Love to the point of invention! This looks and feels great.
I'm an Elixir newbie and wondering if I should start with learning Ash or stick with Liveview until I know more. Any thoughts on what Ash solved for you over Phoenix Liveview?
Ash can be used in conjunction with Phoenix, they aren't mutually exclusive. Ash is really just a framework for modeling your domain(s) and getting a bunch of helpful functionality for free (e.g advanced querying capabilities, pagination, data validations, json+graphql apis, and more) Then you could use those functionalities with phoenix to build a full web app. Or you could use something else other than phoenix, it's up to you :)
Def start with Phoenix first and then try Ash.
I'm actually building open-source SaaS for every vertical and leveraging that to build an interoperable marketplace. I'm nearly done with the restaurant one (alternative to Toast).
This looks great! Love the screenshots, although I'm unable to login to the app. Likely due to the extra traffic this post has gotten. How long did it take you to build this ?
Yeah, trying to pump more memory. It took a few weekends over a long period of time
First time hearing the term micro bakery. Is there an equivalent for restaurants? Like a micro restaurant? Seems like a fun idea for my retired parents.
As someone who struggled with ERPs, this is super-nice and clean!
Thanks!
Looks well thought out. We wrestle with website, real ERP and building Notion connectors for production orders in make to order scenarios so there’s definitely a pain point.
Yeah. Notion is not a bad idea but you'll need to maintain the connections. In fact i did model the idea first in notion
You built your wife?! This robotics stuff has gone too far. Next you will be telling me that you threw your cow over the fence some hay!
Rather offtopic question: what browser are the screenshots taken in? The window chrome looks familiar but I can't put my finger on it.
Looks like Arc Browser.
Yup, that's Arc
Thanks!
Live demo not loading for me - hug of death?
This is unreal. Nice work How long did it take ? I tried to use ash to build a simple app and couldn’t get it to work lol.
I’m an elixir noob
A few weeks (mostly weekends)
You've done a great job!
It's simple, well documented, and uses appealing technologies.
I'm sure your wife's business will take off.
This is really awesome, congrats!
Honestly, well done and thanks for sharing it. I also really appreciate the fact that you included multiple screenshots of the UI, as well as some of the agent plans. Reading the code and project structure, it feels like you put in the work.
That is one good-looking website.
Seems like the live demo website is about to die though.
Yeah, didn't expect so much traffic. Going to pump some RAM
This looks amazing! How long did it take to get to this state?
Few weeks. Elixir + Ash + LiveView is the best productive combination
Login submission hangs forever, guessing too much traffic
awesome and thank you for open sourcing it! Maybe consider renaming it, since there is CraftNote, an app with similar scope in construction business.
This is why technical people make the best spouses
This is incredible work. Can we get a blog post?
Maybe? What would you like to hear about?
Well done Claude, good job!
this is sick, my wife complains about her franchise's mandatory technology portal bs etc. I was thinking I could definitely whip up something better and you've inspired me more.
Incredible