The non-thinking Kimi K2 is on Vertex AI, so it's just a matter of time before it appears there. Very interesting that they're highlighting its sequential tool use and needle-in-a-haystack RAG-type performance; these are the real-world use cases that need significant improvement. Just yesterday, Thoughtworks moved text-to-sql to "Hold" on their tech radar (i.e. they recommend you stop doing it).
Thanks, I didn't realize Thoughtworks was staying so up-to-date w/ this stuff.
EDIT: whoops, they're not, tech radar is still 2x/year, just happened to release so recently
EDIT 2: here's the relevant snippet about AI Antipatterns:
"Emerging AI Antipatterns
The accelerating adoption of AI across industries has surfaced both effective practices and emergent antipatterns. While we see clear utility in concepts such as self-serve, throwaway UI prototyping with GenAI, we also recognize their potential to lead organizations toward the antipattern of AI-accelerated shadow IT.
Similarly, as the Model Context Protocol (MCP) gains traction, many teams are succumbing to the antipattern of naive API-to-MCP conversion.
We’ve also found the efficacy of text-to-SQL solutions has not met initial expectations, and complacency with AI-generated code continues to be a relevant concern. Even within emerging practices such as spec-driven development, we’ve noted the risk of reverting to traditional software-engineering antipatterns — most notably, a bias toward heavy up-front specification and big-bang releases. Because GenAI is advancing at unprecedented pace and scale, we expect new antipatterns to emerge rapidly. Teams should stay vigilant for patterns that appear effective at first but degrade over time and slow feedback, undermine adaptability or obscure accountability."
Vertical: industry specific (functionality that can only be used in one or a few industries). The ultimate case of this would be a completely custom solution a bank uses to do x, with a custom integration to their databases etc.
Horizontal: Can be used by many industries as they all need it (accounting software, communication & marketing etc.) Office 365 or Slack are classic examples.
Utiligize is looking for a Senior Data Engineer to solve one of the most difficult problems in decarbonization: building out power system infrastructure to support huge increases in electricity demand and production.
You will take on development of data pipelines, ETL/ELT processes, setting up task runners, optimizing queries and choosing the right infrastructure to support simulations based on petabytes of data. You will have significant independence in choice of infrastructure, cloud or self hosted, and a large part of the job will be experimenting with new technologies as we solve problems that haven't been solved before.
Desirable skills: Python, SQL and bash. Task runner and job scheduling software like Airflow or Luigi. Experience with Apache Spark, Google Bigquery, Clickhouse, or other columnar or distributed database technologies. Knowledge of spatial or topological data and time series data is great, but not mandatory.
You hold an EU working permit, or have Ukrainian citizenship. Competitive compensation ($100-120k) and equity offered. Email your CV to jobs@utiligize.com
Utiligize | Copenhagen, Denmark | React Engineer, Infra Engineer | Full time | Onsite | https://utiligize.com
Our software helps utilities invest in the right infrastructure to support challenges related to electric vehicle charging, heat pump and solar deployment. Our platform includes a mobile app for data collection and a dashboard for visualisation and control.
We're looking for a React developer to understand customer requirements and develop new features. It is preferable if you also understand Node.js, Postgresql and Typescript.
Having some mapping (GIS) experience with Mapbox or Leaflet is a plus, but it's also something you can learn on the job.
We offer competitive benefits packages (salary, equity, free lunch, 401K, fully subsidized personal healthcare, cell phone allowances, etc). Get in touch at contact@utiligize.com
That's not what section 230 says today, but there's a very interesting debate to be had about what its inevitable replacement should say tomorrow. Ranking posts according to some unexplainable algorithm which includes things like keyword extraction, often "selfishly" to favor engagement, has proven to be far from benign. I think it's quite reasonable to say that as a platform exerts more of this control it should also assume more responsibility. If you're not prepared to take on that responsibility, stick to strict chronological order and/or user defined priorities.
I don't particularly like it when Facebook (for example) buries content from my actual friends and family beneath posts that it thinks might be more engaging. They're usually wrong, BTW; the moment I recognize it as an algorithmic promotion I scroll right past as quick as I can. They certainly shouldn't be showing me stuff from pages and groups I never expressed an interest in. If I want to find new sources I'll ask. If they do those things, they are acting as editors and publishers, and should be treated as such. There are still problems to be solved around groups people have already joined and ads and privacy, but if they'd at least stop pulling every user toward more extreme content - effectively recruiting for the worst of the worst - that would be positive.
Thanks for that clarifying link! I still wonder about this, though. Relevant section (c)(1) says:
>No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.
I'm hung up on the words "provided by". Facebook's algorithm controls what is presented to each user. They are providing a view of some posts, and not others. Facebook is creating the wall for each user, right?
Or would all this still considered moderation, allowing them to do what they want? Section (c)(2)(B) mentions not being liable for allowing users to control what content is accessed, but doesn't mention when the provider makes decisions like this.
At an extreme, could facebook use their secret algorithm to promote all posts saying "stolen election" to all republicans, demote contrary posts, and still claim claim section 230 protection because they didn't create the posts even though they could choose whatever they want to go viral amongst millions of various posts?
If tractors with closed software are such a problem, why doesn't a company build old-school tractors? Heck, you could use a 1980 design, currently selling second-hand for $40-60k, based on expired patents.
Farmers need same-day access to replacement parts, which means a dealer with a full stock of parts within driving range. Belarus/MTZ tractors are low tech designs but have about one dealer in every state.
Unfortunately it's not a mistake. I've worked on quite a few joint projects between the Nordic transmission system operators and this is just one of those areas where it's different in each country. Sweden is usually the strictest.
