If patterns observed in training data usually dominate exhibited behaviors of ML models, then limited training data seems like a good explanation. Of course, transfer learning may be another factor. If abstract-ness is the cause, then English may be a factor. Interesting space :)
Are they bad at TLA+ because of limited TLA+ training data? Or are they bad at writing specifications because of limited specification training data? Or is English a bad specification language?
By precedence, any SM account can make statements like "Grammarly suggested XYZ is bad" until #Grammarly contacts each account to opt out of such statements, right? How are companies running without common sense thinking? 🤦🤦🤦
I feel the same way about OSS but from a different viewpoint: has it become exploitive? Is it harder to monetize for creators? Still gathering thoughts.
Why not combine the review system and the code generation system, and generate code without logical errors? 🤔 It seems like a case of a merchant selling both the poison and its antidote.
"Whatever courts eventually decide about AI reimplementation, the question we need to answer first is not a legal one. It is a social one. Do those who take from the commons owe something back?"
#OSS #copyleft #license #software #GenAI
writings.hongminhee.org/2026/03/lega...
Interesting take on how #software has evolved from a passive tool to an always-on, interrupting, and snooping entity 👏 Also, on how enabling techniques like telemetry and experimentation have affected #softwareengineering 👏
blog.mikeswanson.com/backseat-sof...
Source: "Design Lobster" newsletter
Is "yep dawg, I am aching" the right response? 😁
I've been expecting to see this shift for a while and it is interesting to see it actually starting to happen.
Rich and balanced comparison of the performance of SPA and hypermedia based web app with interesting findings 👏 The post includes limitations/caveats along with balanced when-to-use recommendations 👍
#SPA #web #hypermedia #performance
zweiundeins.gmbh/en/methodolo...
For me, programming in #Elixir is fun, easy, and simple due to pipes. The program is a sequence/flow of data transformations. The same is true in #Clojure (threading macros) and #FSharp.
Awesome analysis about the #human aspect of #Linux #kernel bugs 👏👏👏
Many interesting observations that can be considered by all #software #engineering teams.
pebblebed.com/blog/kernel-...
Seems similar to Model Driven Development (MDD): go from a model (aka spec of a system) to an implementation. Trusted translation/compilation and traceability were big issues with MDD. With proposed regeneration, I think backward compatible evolution will be an additional issue.
The skill to know when not to act and then to not act is often looked down upon; more so, when "bias for action" is all the rage.
"The only solution is to educate our children about life with digital abundance. Throwing them into the deep end when they’re 16 or 18 is too late." -- blog.system76.com/post/system7... by @system76.bsky.social
I am reminded of bsky.app/profile/alex.... Was Internet an information discontinuity?
TIL eventually all paid competitions at #kaggle will require identity verification to ensure every user is a live human with a single account 🤦 Verification is done by uploading selfies to the Persona company 🤦
#security #privacy
www.kaggle.com/contact#/acc...
I am convinced we are on the verge of the first "AI agent worm". This looks like the closest hint of it, though it isn't it quite itself: an attack on a PR agent that got it to set up to install openclaw with full access on 4k machines grith.ai/blog/clineje...
👏 The sentence following what @carnage4life.bsky.social quoted paints a fuller picture: "Personally, I saw very limited gains in productivity from AI, nothing nearly profound to justify tossing out half of the company's workforce along with institutional knowledge and expertise."
My kid's teacher objected to a write-up suspecting it may have been generated by AI. The trigger was the use of richer than usual vocabulary, which resulted from discussing and reviewing the write-up with me. How are students supposed to learn without discussion and feedback?
What could go wrong with AI? 🤦
#AI #technology #LLM
So, can companies continuously capture/build and share their #institutional #knowledge across their workforce? If so, #LLMs can help one of the most frustrating aspects of corporate work 👏 Of course, there will be security risks (e.g., need-to-know info) that are worth tackling given the benefits.
Oh, a new way to deal with Monday blues or scratch one's masochistic itch 🤣 Why? Well, if you are seeking more pain, then may I suggest migrating by using Outlook's web interface via a non-MS browser on Linux?
Nice overview of actor-based approach to concurrency -- isolated processes, messages, and supervision -- supported by BEAM and OTP.
In contrast to leadership exhibited by the likes of Gandhi and MLK, the modern day spiel about leadership is about building a team (which I think is more about management which has its place) and not about pursuing and inspiring folks about worthy causes.
Can an agent add a "star" to a repo, say, if it uses the repo? I would not read too much into such metrics.
Given the recent geopolitical and economic rifts between USA and EU and Canada, I wonder "how will the EU nations and Canada respond if USA were to request help in the latest middle east conflict (like in the 2000s)?" Let's hope we don't have to answer this question🤞
A great overview of the evolution of #container technology via #Docker. It also provides a quick peek into the various technology bits that enable containers work.
If you want to learn more about how containers work, then consider starting with wizardzines.com/zines/contai... by @b0rk.jvns.ca.
As "cloning" software becomes easier with LLM-based SE (which works better with oracles, e.g., tests), can tests serve as the new moat? They can but only until we figure out efficient ways to generate effective tests.
saewitz.com/tests-are-th...
#oss #softwaredevelopment #LLM
You can. Just turn off the computer for a weekend 😛 Jokes aside, struggles to connect devices and Internet seems to be the bane of life with computers.
Neat idea that leverage randomness in natural phenomena -- randomness in a tear -- to improve security.
#technology #security
news.mit.edu/2026/chip-pr...
