Never heard of it, but seems highly relevant and well thought out. Our managers always stop at the vision and mission part, stuff some buzzwords into the soup and then offload the implementation to the team or individual. That success is following a holistic chain of events and needs attention beyond initial flag waving is not self explanatory. I will look into this more deeply and will bring it up eventually.
I'm downloading an entire forum now. I did not have much luck with tools like HTTrack. LLMs are very useful to create a custom scraper. I indexed the forum first (got URLs and titles of every thread) and then I supply the index to a second program that downloads the threads. I even managed to get third party embedded image assets pulled in, just one extra description in the spec before feeding it to the LLM. I keep getting blocked after some 200 fetches, but I have a dynamic IP.
I have been using HTTrack is cases where I want a browsable local version of any site (not just a page)! Most of the applications mentioned here as comments only allow for saving a single page. But afaik, only HTTrack allowed me to save a local version of the website and not just a webpage. The “depth” argument allows me to do that when downloading a site through CLI
I am in an enterprise which experiments with ontologies and knowledge graphs since its hype heydays ca. 15 years ago. It never took off, good ol’ relational databases and its modern day implementations in the cloud still rule the data landscape. Many smart consultants tried their best and we spend many millions over time into it. I would love to read more success stories, but they seem to be rare beyond academic exercises.
95% of sparql, cypher, and gremlin can be lowered to sql anyway. I rolled my own sql lowering system for duckdb over iceberg which lets me do everything in that blog post but with no bespoke data layer.
Strikingly similar story here. Ontologies and knowledge graphs seemingly make a comeback because it is argued (with little evidence, from what I can tell) that it helps LLMs make sense of the dataset. I see no practical difference or benefit vs. providing the date model as DDL or via other means to represent the ER graph. I'd like very much to be convinced otherwise or at least read a rationale about the alleged differences.
Thanks for the link. I did go through it despite the annoying style.
It postulates that semantic models delivered in areas where SQL couldn't, that what hinders their broader adoption is the user-interface/querying layer (i.e. too few people adopting SPARQL), and suggests that a new dialect of SQL could solve that, especially because LLMs are so good at producing SQL.
> Knowledge graphs and semantic models promised something enterprises desperately needed: governed meaning, reusable business logic, and explicit relationships across complex data landscapes. And technically, they delivered.
For me that's the main argument: what is the value proposition for knowledge graphs and semantic models when you already have an established relational datamodel? Some argue that the graph's fluid schema brings flexibility benefits, but looking at the problem in its entirety (considering the effort to define and formalise the model, the effort to store content, the effort to retrieve it, deliberately keeping the dev/sysops and tooling out of the way), what you spare at formalisation/insertion stages, you pay back at retrieval (or, if you don't validate or constrain your inputs, expect surprises as an output).
That's where I stand right now:
- convince me that relational models can't express as powerful constructs and concepts as semantic ones
- convince me that "fluidity" in the model don't come with more troubles down the road than benefits
> Knowledge graphs exist. […] And almost nobody queries them directly. Not because the models are wrong. Not because the semantics aren’t useful. But because accessing them requires SPARQL
Or, guess what, a similar outcome can be explained differently: perhaps it's not SPARQL that's too difficult to grasp. Perhaps the alleged benefits of semantic models just do not materialise in practice, and there is no point to engage with SPARQL to begin with because there's no payoff at the end of it. I think the data processing ecosystem is sufficiently vivid and diverse to demonstrate that users have no problem embracing new technologies and languages as the need arise (sed/grep/…, hadoop/spark, pandas/polars/duckdb, decades of ETLs in graphical/textual forms, …)
Then the article is full of weird takes, especially within the "Hidden Cost" section, that I don't feel like commenting. Nor why their "SQL with semantic ontologies" is just syntactic sugar to save your editor from writing joins on your behalf?
I mean, doesn't this exact statement also apply to the adoption of AI and all the winters that have come before this wave?
Also as someone doing my own implementation at the moment, I expect that there is a big relational database backing the MELT store, just the graph aspect is functionally a big RAG interface on top of it acting as a search cache, not itself the data store.
That’s good. I wonder how much longer it will take for countries to make their energy supply dependent on fossil fuels or nuclear power, thereby becoming pawns of unpredictable or hostile regimes and being helplessly at the mercy of geopolitics. It’s so baffling that many countries, even those with plenty of sun and wind, still prefer to rely on fossil fuels rather than make life a little easier for future generations.
And no, there will be better solutions for base load as well. Battery storage, for example. Sure, it’s a transformation, but those who refuse to think for the future can see the consequences of stubbornness and backward-looking attitudes today.
Contrary to their name, "rare earth metals" aren't actually rare, and for a lot of the renewables stack they aren't strictly necessary.
Take lithium batteries, for example. They were attractive for early BEVs because the consumer electronics market already developed an industry for them, and they were attractive for early grid storage because the BEVs scaled that industry up. But that doesn't mean you have to use lithium for grid storage, and when push comes to shove we could also use sodium-ion batteries, or iron redox flow batteries.
If your country no longer has access to iron and salt, I reckon you have bigger things to worry about.
Didn’t say they were rare I said they weren’t distributed equally which ends up creating geopolitical tensions and while newer chemistries might change the equation that doesn’t seem like what is going to happen at least in the short term.
Renewable energy is really expensive. [About 4x more expensive than nuclear, and 14x more expensive than gas.](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4028640) People mistakenly think that because wind and sun are free, renewable energy generation is cheap.
Your four year old paper shows wind+solar costs as on par with nuclear in the final (LFSCOE95) model in the ERCOT market before factoring in dropping battery farm costs.
Other papers exist with a different opinion (eg: Australia's more recent CSIRO energy report finds nuclear not providing bang for the buck in the Australian market)
Further, despite claiming Level Full System Costs, there's no factoring in for externalities - eg: climate costs of dragging out fossil fuels while waiting for nuclear to come online, political costs of being steadily reliant on outside countries for fuel and supply (cost of gas to non gas nations, etc).
There are other issues - it's a paper with an opinion - not the foundational rock you want to base strong assertions on, as you did.
An odd criticism about the paper being four years old. I assure you, the laws of physics both existed back then, and we had a good understanding of how electricity is generated. Solar and wind costs have reduced since then but nowhere close to bridging the gap. Feel free to provide some evidence if that’s what you’re implying. Let’s take a look at CSIRO report, because I don’t think you have.
The ERCOT LFSCOE-95 scenario assumes another generation source supplies the remaining 5% of demand. Even then, wind and solar cost $97/MWh versus $32 for gas. And in Germany, the same model gives $220/MWh for wind and solar, $90 for nuclear and $31 for gas. So even your preferred scenario doesn’t support the conclusion you’re drawing.
Dismissing a peer-reviewed paper published in Energy because it’s four years old is particularly weak when you haven’t demonstrated which assumptions are invalid or how newer data overturn its conclusions.
And here’s a May 2026 peer-reviewed ERCOT study using detailed hourly dispatch modelling. It found that a more renewables-heavy electricity mix increased average costs by 12.1% relative to the alternative:
As for externalities, those are legitimate additional costs to quantify, not an excuse to pretend that intermittency, storage, backup capacity and grid integration are free.
Calling a quantitative study an “opinion” while offering little more than your own opinion is rather ironic.
Your data is way out of date. Renewables, particularly solar+battery, are incredibly attractive from an economic perspective. That’s why the build-out has accelerated so much. It’s not government hippies (I wish—have you seen the state of world politics lately?), it’s cold hard capitalism.
There is a spectrum between using AI as tool and for automation. As humans are lazy, they always go more for the automation end if it accessible. And the human skill at stakes is brain capability. So yes, I agree: AI looks like an apocalypse for human thinking.
Yeah, using it a bunch with local LLMs. It's fast (small system prompt), focused on the basics, made around simple choices, and possibility to call it in scripts.
So far I have very positive experiences, the results are generally better (and not just faster) than in other harnesses with the same models. I've done the same (user) prompts for tasks in a range of them, and compare the outcome to eval between them. Simple (harness) beats complicated one many times. I do wish it could use MCP though, but not as often as I think, and CLI tool usage often compensates on this front with little extra effort)
That is now my usuall go-to harness for things usually. Still figuring out the edges though (when we get off the happy path, like models reaching their max context, or such...)
It's very light, recommed you try it out, and see if it fits your toolbox.
Just tried. And wow!! I opened my mac's activity monitor to see how much load it was putting on my system, and i had to scroll all the way down ... unlike those current harnesses that sit right on top (sorted by used cpu, descending).
And it worked. I'm switching to this unless i find a reason not to.
it's minimal and somewhat opinionated, not integrated well with memory subsystems (although that's kind of the point)
my only real criticism is that last I played with it it deals with slow provider responses very poorly , so lots of locally hosted llms require a lot of tuning to get it to handle streaming/non-streaming/timers well.
way less minimal (but still pretty minimal) I prefer Pi; but different markets I think.
hax fit my openbsd laptop well, espescially since Pi and related derivatives are kind of un-spokenly incompatible with the BSDs.
(author passing by) Could you please report the issues you've experienced on the repo? I did quite a bit of testing myself on local models (Qwen, Ornith, Gemma, Muse Glimmer) with llama.cpp, and they all worked fine. Details would be helpful.
Also one brief note for Codex users specifically. hax can work out-of-the-box as codex-cli drop-in replacement (since it can use the already authenticated codex token), so it's super easy to try it - basically it will be a zero-config experience.
Firstly thanks for writing it and making it available. The problem I had when I tried it is it would consistently blow out the small context window of my local qwen and although there was seemingly a config knob I could use to prevent that, seemingly that didn’t do anything.
That landing page is hard to look at, let alone read. Have LLMs really not yet learned to write webpages that read the sytem theme and look that way? Must be hard.
I wonder what is the value of such vibe coded programs. It can't be the implementation or engineering aspect at all, when generated by AI. Is it the idea? Should it be a tribute to Craig Reynolds in this case?
The first commits are from September 2025, when agents were far less capable than now. Later I decided to use AI to implement some features.
As for the value: exploring ideas, techniques and features that would have been impossible to explore completely by hand (if you have a life), and a final artifact that is cool and pleasant to watch in your terminal.
I guess all these programs are just ideas now, instead of a feat of engineering?
So more like a comment or idea shared at a social gathering, that anyone else might build on with their own in conversation. rather than the prior situation of code, where it is as if we were in wonder at the very fact of speech itself
At least that's my optimistic take. Though many might find the same thing a bummer.
There will still be poets, and those who admire them
Craig Reynolds worked for an AI company named Symbolics, for Christ's sake! ;)
So what was the value in 1986 of having a Symbolics Lisp Machine calculate the flocking behavior and render the graphics instead of working out all the animation with a slide rule and graph paper then drawing each frame by hand?
Cheating by using a powerful AI programming language like Lisp and expensive Lisp Machine hardware and Genera AI programming environment to automatically perform all the hard work sucked all the fun and artistic expression and human touch out of it. Think of how many pencil pushing number crunchers and pen and ink animators Boids have put out of work! /s
If you're going to be an anti-AI luddite, then be consistent and go all the way.
reply