Blogs

The climate constraints are currently not compatible with the ongoing arm race on large neural networks models. The training seems kinda OK, but the inferences… and it’s currently just rolled out as shiny gadgets. This really need to be rethought.

https://limited.systems/articles/climate-cost-of-ai-revolution/

The Unpredictable Abilities Emerging From Large AI Models | Quanta Magazine

Tags: tech, ai, gpt, complexity, emergence

Now, this is interesting research. With all that complexity, emergence is bound to happen. There’s a chance to explain how and why. The links with the training data quality and the prompts themselves are interesting. It also explains a lot of the uncertainty.

https://www.quantamagazine.org/the-unpredictable-abilities-emerging-from-large-ai-models-20230316/

Hallucinations Could Blunt ChatGPT’s Success - IEEE Spectrum

Tags: tech, ai, machine-learning, gpt

Now this is a properly balanced piece which looks beyond the hype. Usable yes, if hallucinations don’t have a high impact. Can the hallucinations be solved? To be seen, I personally have my doubts with the current architecture… at least banking it all on human feedback is being very naive about the scale of the task.

https://spectrum.ieee.org/ai-hallucination

Google and Microsoft’s chatbots are already citing one another in a misinformation shitshow - The Verge

Tags: tech, ai, gpt, google, microsoft, criticism, fake

Training sets are obviously already contaminated… now it’ll be a race of hiding such mistake under the carpet with human interventions. That’ll be a boon for misinformation. That’s what we get for a useless large models arm race.

https://www.theverge.com/2023/3/22/23651564/google-microsoft-bard-bing-chatbots-misinformation

ChatGPT Gets Its “Wolfram Superpowers”!—Stephen Wolfram Writings

Tags: tech, ai, machine-learning, gpt, computation

Now, this starts to become interesting. This is a first example of trying to plug symbolic and sub-symbolic approaches together in the wild. This highlights some limitations of this particular (quite a bit rough) approach, we’ll see how far that can go before another finer approach is needed.

https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/

Epic’s new motion-capture animation tech has to be seen to be believed | Ars Technica

Tags: tech, ai, 3d, animation

Now this is a truly impressive technology! This will make facial motion capture a really smoother process.

https://arstechnica.com/gaming/2023/03/epics-new-motion-capture-animation-tech-has-to-be-seen-to-be-believed/

Image Codec Comparison

Tags: tech, photography, codec, art

Interesting benchmark, this seems to point toward AVIF and JPEG-XL as two great codecs for pictures.

https://giannirosato.com/blog/post/image-comparison/

ThumbHash: A very compact representation of an image placeholder

Tags: tech, graphics, frontend, hash

Interesting algorithm for generating image placeholders.

https://evanw.github.io/thumbhash/

TinyVG

Tags: tech, graphics, vector

Looks like an interesting vector image format. Let’s see if it gets some buy in.

https://tinyvg.tech/

Modern Font Stacks

Tags: tech, frontend, web, fonts

Very nice approach to avoid the font bloating on the web. I’m slightly concerned about the maintenance over time but at least it has proper fallbacks and the fonts used seem widespread enough (for now).

https://modernfontstacks.com/

Introducing Ares - The Fastest Way to Decode Anything

Tags: tech, tools, cryptography, encodings

Still a bit young but looks like an interesting and fast tool to decode random data.

https://skerritt.blog/introducing-ares/

Envio is a command-line tool that simplifies the management of environment variables

Tags: tech, tools, command-line

This looks interesting, I especially like the fact that it’s easily encrypted, definitely a good thing regarding secrets. Now I wonder if that’s easy to couple with direnv…

https://github.com/humblepenguinn/envio

Why I Will Never Use Alpine Linux Ever Again | Martin Heinz | Personal Website & Blog

Tags: tech, docker

This is a very good point, it’s good to reduce containers size. Be careful of the cost though.

https://martinheinz.dev/blog/92

Laurence Tratt: How Big Should a Programming Language Be?

Tags: tech, programming, language

Interesting musing about a language size and how it evolves over time. There’s clearly tension between making it too big and keeping it relevant to modern uses.

https://tratt.net/laurie/blog/2023/how_big_should_a_programming_language_be.html

Why people misuse inheritance | Thoughts, solicited and otherwise

Tags: tech, object-oriented, programming

A good point… everybody should know at this point that delegation should be favor. So why do we keep turning to inheritance even in cases we shouldn’t? Convenience and writing less code mainly. Unfortunately that leads to bugs more often than not.

https://solicited-thoughts.bearblog.dev/why-people-misuse-inheritance/

System design and the cost of architectural complexity

Tags: tech, architecture, complexity, hr, bug, productivity

It’s been a while since I dived into reading a Ph.D thesis… I bumped into that one through an article which was trying to summarize it but I wasn’t super happy with it. That’s why I decided to go to the source.

It’s an interesting read, it has the benefit of making a clear difference between complicated and complex from the get go, which is welcome (generally a good sign for me). If you want the tl;dr it’s at the end of page 16: “we found that differences in architectural complexity accounted for differences in developer productivity of 50%, three-fold differences in defect density, and order-of-magnitude differences in staff turnover”.

Note the last point about the staff turnover should be taken with a grain of salt though. It is well explained in the limitations of the study, being a lot in the high complexity areas of the code can also be a sign of higher skills and thus more job opportunities.

Anyway, I think we all suspected some link between complexity and productivity but I always wondered how much. Seeing how the study was done it’s definitely not an absolute answer (very thorough and precise, even historical data taken into account over several releases… but in a single company). Still the value is in at last giving us some rough numbers on how far the impacts can go. Thus, the scale of those impacts are potentially huge.

Maybe it’s time to stop trying to find rockstar developers or mythical 10x developers (common “leprechauns” of our industry)… Let’s focus on tackling undue or uncontrolled architectural and code complexity instead, shall we? Even better if that’s done through the use of documented patterns when applicable.

Interestingly, the literature review part gives a few clues about why there is under-investment in architecture in general, or reworking the architecture on long term project. It’s unclear to organizations the costs of the undue complexity will carry. It’s exactly what this thesis tries to shed light on (see tl;dr above).

Also, it’s interesting to see confirmed that the perception of the architectural complexity we have is often wrong when looking at parts in isolation. The relationships need to be transitively mapped to start to grasp the presence of architectural complexity. That’s why only coordinated efforts can tackle it, it’s almost impossible to tackle for a single developer.

Of course I’d advise reading it in full, that requires investing some time into it though.

Very stimulating, I’d like to apply some of those tools on projects in the wild but I’m not sure there are ready made tools available. Also I’m wondering what we would find if I’d reuse some of those in ComDaAn to work on temporality of changes rather than dependencies. I think this could give interesting insights.

https://dspace.mit.edu/handle/1721.¹⁄₇₉₅₅₁

Go slow to move fast - Thoughts on software and technical leadership

Tags: tech, project-management, technical-debt

Technical debt was an interesting metaphor to kickstart the conversation but has been overused. It can still be useful, especially with the proposed approach here to make it intentional and explicit. This can be factored in how to drive the project.

https://jordankaye.dev/posts/go-slow-move-fast/

Project Management for Software Engineers | Kevin Sookocheff

Tags: tech, project-management

This is a good summary of the most important points in the PMI body of knowledge. If you dabble in project management it’s worth looking at it.

https://sookocheff.com/post/engineering-management/project-management-for-software-engineers/

I’ve been employed in tech for years, but I’ve almost never worked – Emmanuel Maggiori

Tags: tech, business, agile, criticism

OK, not a perfect article, I think there are a couple of blind spots in the reasoning (I doubt all the estimates were as systematically bloated as presented here). Still, it’s another interesting account of the problems created by the cargo cult agile. It indeed seems to resonate with the fact that the tech sector is very hype driven. A lot of useless work then ensues.

https://emaggiori.com/employed-in-tech-for-years-but-almost-never-worked/

Why Construction Projects Always Go Over Budget — Practical Engineering

Tags: tech, engineering, estimates

For all we like to point out the software industry for blowing up estimates and budgets… it’s not a unique phenomenon, civil engineering is also struggling with it. This is a good reminder.

https://practical.engineering/blog/2023/3/21/why-construction-projects-always-go-over-budget

Culture Viruses | Stay SaaSy

Tags: management, hr, culture

Interesting way to frame the potential problems around organizational culture. This indeed influence behaviors quite a bit so should be in check. It also shows it’s a complicated problem you don’t want to overdo it, freeze the culture in place, and see it used mainly for blaming… it’d effectively turn into a cult.

https://staysaasy.com/business/2023/03/01/culture-viruses.html

Bye for now!