I think the best place to put barriers in place is at the mcp / tool layer. The email inbox mcp should have guardrails to prevent damage. Those guardrails could be fine grained permissions, but could also be an adversarial model dedicated to prevent misuse.
That's the point of the loop, (the prompt is in another comment) start with a fresh context at every step, read the whole code base, and do one thing at a time.
Two important part that has been left out from the article is 1) service code size, our services are small enough to fit in a context + leave room for implementation of the change. If this is not the case you need to scope it down from 'read the whole service'.
The other part is that our services interact with http apis specified as openapi yaml specs, and the refactoring hopefully doesn't alter their behaviour and specifications. If it was internal apis or libraries where the spec are part of the code that would potentially be touched by the reafctoring I would be less at ease with this kind of approach
The service also have close to 100% test coverage, and this is still essential as the models still do mistakes that wouldn't be caught without them
The architectural patterns are similar in go. The part of the prompt that contains the refactoring concerns that I wanted to fix are specific to this go project. You can very well add what you just explained and not only will it follow it, it will cleanup the parts when it isn't done. You don't need to fully explain the concept as it probably nows them well, just mentionning the concept you want to fix is enough.
In my experience the latest model (Opus 4.6 in this case) are perfectly able to do senior stuff. It's just that they don't do it from the get go, as they will give you the naive junior dev solution as a first draft. But then you can iterate on refactoring later on
> …You don't need to fully explain the
> concept as it probably nows them well…
Unsurprisingly, many would disagree [1]…
> 1 Establish a Clear Vision
…
You have experienced the world, and you want to work together with a system that has no experience in this world you live in. Every decision in your project that you don’t take and document will be taken for you by the AI…
Unfortunately not, it's private company code. But I can share the prompt I used for the refactoring:
- 1. Read the whole code of the repository.
- 2. Read the TASKS.md file if it exists.
- 2.1. If it exists and is not empty, pick a refactoring task from the list. Choose the most appropriate.
- 2.1.1. Refactor the code according to the task description.
- 2.1.2. Commit the changes to git.
- 2.1.3 Remove the task from TASKS.md
- 2.1.3. You are done.
- 2.2. If it doesn't exist or is empty:
- 2.2.1. Identify the parts of the code that could be refactored, following the following principles
- A class should have a single responsability
- The dependencies of the class should be mockable and injected at class instanciation
- Repeated code should be factored into functions
- Files shouldn't be longer than 1.5K lines
- 2.2.2: If using the previous insights you think there is valuable refactoring work to be done:
- 2.2.2.1 Write a list of refactoring tasks in TASKS.md
- 2.2.2.2: You are done.
- 2.2.3: If there is no more refactoring to be done, notify me with 'say "I am done with refactoring"'
The TASKS.md file will be created and filled by the model. The prompt needs to be run repeatedly in a loop until it decides there's nothing to be done anymore.
I for one found this event really sad. It's like the OSS community has rejected the past 5 years of software and technological changes and now choses to live in a retro computing bubble.
We're in 2026, hardware is made in dark factories in shenzhen in fully automated assembly lines by the million of units. Software is written using LLMs hosted in gigantic datacenters. Millions of people are now writing their own software with vibe coding platforms from their phones
What is the FOSDEM community's answer to the real concerns that these changes pose ? Let's hand solder raspberry pis ! let's self host LLMS from 2 years ago on FreeBSD ! Look, i can run wasn linux on this risc-v cpu !
These takes are completely out of touch with reality, no wonder that nobody younger than 40 was attending the conference. The next generation is doing something else and rightly so.
> What is the FOSDEM community's answer to the real concerns that these changes pose ? Let's hand solder raspberry pis ! let's self host LLMS from 2 years ago on FreeBSD ! Look, i can run wasn linux on this risc-v cpu !
Maybe, just maybe, they're having fun? FOSS is not only about corporate open-source, but also genuine curiosity. Both can have their place.
> no wonder that nobody younger than 40 was attending the conference. The next generation is doing something else and rightly so.
I saw a lot of students at FOSDEM, attending, presenting and helping the at organization.
OSS original point wasn't just to have fun among nerds, but to have a real impact on the world. Fun and curiosity is fine, but there's a line where it becomes a tech themed larp event, and FOSDEM is trending towards the later.
Corporate Open Source should have its place at FOSDEM. The linux dinosaur companies such as Redhat are still there. But what about the new ones ? What about Mistral, Odoo ? Even the 'evil' ones such as facebook, github, etc, aren't they contributing a lot of open software ? Aren't they more relevant than let's say Olimex ?
There were some students, yes but the attendence is growing old, and the chit chat is more about 'remember this and that' than 'we're building the future'
I did find devrooms, stands, and main track talks on this.
> Even the 'evil' ones such as facebook, github, etc, aren't they contributing a lot of open software
Google is one of the two lead financial sponsors of FOSDEM (the other being RedHat). So clearly, there doesn't seem to be any restriction or judgement by the organizers on whether BigTech is 'evil'. I get that Jack Dorsey (of Square/Block) withdrew his main track talk the year prior, which was unfortunate.
> the chit chat is more about 'remember this and that' than 'we're building the future'
Well, for better or for worse, FOSDEM is not exclusively a tech start-up event.
It doesn't to me at all, it is mainly focused on self hosting llms, which is a complete deadend. It just isn't feasible to self host the useful models, the hardware requirements are just too big.
The current topic of focus around AI are: how to adapt development practice to agentic coding, agent harness, agent orchestration, mcp integrations, etc.
I guess there is some unease in the oss community to rely on large companies to run and host the models. But this isn't entirely new, we also relied on big companies to manufacture our computers. It's just the way it is.
> Well, for better or for worse, FOSDEM is not a tech start-up event.
It is weird, there are a lot of startups present, look at all the stands showcasing projects. Aren't those startups ? What I noticed is that they are usually funded by public grants rather than VCs. I am not sure why this is the case.
I met several people self–hosting LLMs including a man from Tenstorrent demonstrating their accelerator card.
> What I noticed is that they are usually funded by public grants rather than VCs. I am not sure why this is the case.
American VC culture derives from America's privileged position in the global financial system. It can't be replicated by a country that doesn't have a ton of money floating around looking for investments.
Second reason: nobody likes what American VC culture created, so they don't want it to be replicated. Government grant funding can make decisions on axes other than profitability.
Public grants are nice but they have couple of shortcomings and which is why they can get you only that far. They are normally low in capital, the execution is really slow (couple of months to one year), and larger grants too involve politics. The process is too formal (inflexible and too time-consuming) and also quite discriminating to individuals/small-groups who do have the big ideas but are not running the business already (I mean how can they). Proposal evaluation also has its own shortcomings - there's very little incentive for the actual experts to join the evaluation process (it's paid pennies) and generally speaking this leads to another chicken&egg problem - you're presenting something novel to the pool of people who might not have the capacity to understand the idea - neither the vision nor execution.
That said, I am not attracted to the VC culture but their process delivers the value which creates successful companies.
NLNET is always coming up at FOSDEM. Since they have a decent track record of issuing grants, the EU delegates them some money to use in their own less bureaucratic granting process. They call this "cascade funding". NLNET has funded a lot of random individual projects you can find on their website. Nominally, your proposal must have something to do with their goals.
This year there is more emphasis on bringing complete solutions to market. Previously they were funding much more experimentation.
> This year there is more emphasis on bringing complete solutions to market. Previously they were funding much more experimentation.
That's the step in the right direction however there's what I believe is a major issue with the NLNET scheme - there is no fastrack possibility for really great ideas with very potent market impact - you have to spend (lose) ~year to prove your idea is worthy by applying to Zero Commons or similar grant instead of just getting the 200-500k to really get the project hitting the ground.
One year is exceptionally long period in tech, and if the idea is right, you need to have all the resources to execute it - working solely on the project for the whole year for 50,000 EUR is simply not the strategy that can work out in a highly competitive (world) space.
How should they know your project is worth investing 500k? I heard they've got 3x8M, per year I presume, so 500k is a huge chunk of that. Everyone thinks their project is worth 500k, what makes yours different from the rest?
> How should they know your project is worth investing 500k? I heard they've got 3x8M, per year I presume, so 500k is a huge chunk of that. Everyone thinks their project is worth 500k, what makes yours different from the rest?
Well that's the job of VCs, that's what they're expert at.
There's also another model where established industrial communities set up research centers to fund projects that might help their common problems.
Yes, many might believe that their project is worth more than it really is but in my proposition authors of the idea are not the ones who get to decide that but people from NLNET or whatever grant. What I am saying is that currently there is no such process at all and this is a foundational problem with the way how these grants are working.
I guess not. Can you specify it in more concrete terms? They're not just buying your project for an arbitrary price, or VC-investing, they're paying your living costs and hosting costs while you create a donation to the public good, that's how grants work.
They give you 500k if you have a really good reason why you need that... and why the result is worth it... and why your project is more worth it than all the other several projects, combined, they could spend the same 500k on. Most of them are one or two people's living cost for 6 months to a year or so.
I genuinely hope you're a bot. If you're not then please consider being respectful in your conversations and address the question being asked rather than moving goalposts - it is extremely annoying. If you're out of your arguments, learn to say "I don't know".
And I also do hope, if you're a human, that you're not sitting anywhere close to decision making committee be it in NLNET or any other grant program because if you do, it fits into the (terrible) narrative of software market in the EU.
Ad hominem does not apply to bots or trolls and you're one of those two. And I'm not sure what was ad hominem about my response. You're the one being ignorant here
I think the one out of touch is you. Do you really think that depending on proprietary third party services is in the spirit of open source software? You didn't just say people should depend on proprietary models, whose output could at least be considered open source, you're talking about "vibe coding platforms" like lovable, which contain an unavoidable proprietary infrastructure component.
You're also engaging in historical revisionism. "5 years ago" means you're expecting everyone to have jumped on Github Copilot on day one or else they're behind. LLM assisted software development only really took off in the last three years and even then you were still a trailblazer.
Iirc. Jack was accepted by the organizers but pressured out by the community.
Also, about github: Had a chat with the Gitlab chap doing the Git talk in the main track. Apparently they dialed back their involvement with upstream git quite a bit.
Github is currently providing a lot of infra gratis (thanks!) but is at best neutral to code and community.
LLM turds will degrade more and more over time. The newbies will become even dumber and people with actual knowledge will be several steps over anthing else trained to copy and paste from bullshitters.
Just like how novel reading, newspapers, music, rock music, the internet, computer games all made everyone super dumb and the world ended.
I am definitely sure that this time you are right and the world will actually end this time, everyones going to end up way dumber than you, or if they dont, you will still pretend it happened and be smug about it on the internet.
Mobile computer games compared to the ones from the 80's, 90's, and early 00's? For sure. Ditto with tons of best sellers, which are very dull and dumbed down compared to what I could read in Spanish "Dollar Stores" like shops in Spain such as the books Clarke/PKD and Orwell novels for very little, less than what $2 were in Pesetas (pre Euro Spanish currency) in late 90's.
Nowadays you have AI generated books which are very surpar even if they look an impressive exercise of LLM's.
I saw someones profile on goodreads, she was putting away 100 - 200 of those cheap supernatural romance books per year. Reading at that level isnt dumb, the books arent making her stupid. That I consider the books stupid isnt ending the world.
Funny you mention that, because apparently the worst "The Shadow" and "Doc Savage" stories are coming from James Patterson dressed up as bestsellers. And I am a massive massive fan of their pulp outings.
Would you mind showing some compassion for the unintelligent, it's none of their fault.. It's incentives, sociodynamics and manifold other preconditions that bring about these symptoms.
One of these is AI proliferation. GP is not merely virtue signalling — GP is directly contributing to raising average intelligence by discouraging reliance on AI.
Even if well intended, isn't that a matter of reach? whom of the aforementioned vulnerable individuals will ever cross the HN comments section? How many of the HN regulars will not already be weary of any and all LLM outputs?
Have an AI track focused on how to use AI to improve OSS software instead of how to run LLama on RISC-V.
Invite the open source developpers behind popular OSS AI frameworks such as opencode, etc.
Invite talkers from large companies that produce open source software and models such as Mistral.AI
Invite talkers from companies that run OSS LLMs at scale such as groq
Invite the people who build drones in Ukraine, (probably the most succesful open hardware story to date). Have drone building workshops / drone piloting stands
But the AI talked about 3 years ago are not the same kind of projects that are being talked about today. It's rapidly evolving. There is a lot of new scope available for new open source projects to take. This isn't open source reimplementing the same thing for 3 years straight.
That's certainly what the companies who stand to profit want us to believe. But I don't think it is. I think it's yet another hype storm without any actual utility to back it up.
> let's self host LLMS from 2 years ago on FreeBSD
Sounds brilliant. 2024 LLMs were great, and hosted in gigantic datacenters. South Park did an episode on ChatGPT in 2023 to assist with the timescales.
If we can run Free LLMs on a desktop PC that's competitive with 2024 a massive win.
Unless of course you believe that AI will continue to increase exponentially, in which case we're heading for such a different world in 10-15 years that it's meaningless to worry about it and you might as well have fun before the AI decides to kill us all
Remember this is just one conference. This is not a top–down imposition of a new European world order. While the amount of policy discussion is quite high — out of necessity now — and EU representatives attend — this event is still a lot of developers coming together.
This is also why the EU has a hard time. America prefers the capitalist model. EU is trying to find ways encourage individual developers each with their own interests to contribute to shared European goals (which the devs also share but might not find interesting), and they're trying to do this without directly saying "you there, go and do this", and that's harder than herding cats.
For instance they do it through NLNET funding. Devs can propose projects and the ones the EU wants to fund will receive funding. It's a roundabout way to do things. Let's see if it pays off.
From what I gather, this is sort of what happened and why this was even posted in the first place. The models were able to immediately detect a change in their internal state before answering anything.
By funding AI slop, you're funding AI slop, not AI research, or, quote, "drive adoption of AI across strategic and public sectors including healthcare, pharmaceuticals, energy, mobility, manufacturing, construction, agri-food, defence, communications and culture"
> research works on grasping at the possible but not quite knowing how
If these are public money, you want to reduce the blind grasping
> Slop has stumbled into accidental success on a number of occasions.
So, show me these occasions where AI slop led to "transformative potential by driving adoption of AI across strategic and public sectors including healthcare, pharmaceuticals, energy, mobility, manufacturing, construction, agri-food, defence, communications and culture."
