nrvn

I can't "shake off" the feeling that this whole MCP/LLM thing is moving in the wrong if not the opposite direction. Up until recently we have been dealing with (or striving to build) deterministic systems in the sense that the output of such systems is expected to be the same given the same input. LLMs with all respect to them behave on a completely opposite premise. There is zero guarantee a given LLM will respond with the same output to the same exact "prompt". Which is OK because that's how natural human languages work and LLMs are perfectly trained to mimic human language.

But now we have to contain all the relevant emerging threats via teaching the LLM to translate user queries from natural language to some intermediate structured yet non-deterministic representation(subset of Python in the case of CaMeL), and validate the generated code using the conventional methods (deterministic systems, i.e. CaMeL interpreter) against pre-defined policies. Which is fine on paper but every new component (Q-LLM, interpreter, policies, policy engine) will have its own bouquet of threat vectors to be assessed and addressed.

The idea of some "magic" system translating natural language query into series of commands is nice. But this is one of those moments I am afraid I would prefer a "faster horse" especially for the likes of sending emails and organizing my music collection...

from Go documentation[0]:

> Committing profiles directly in the source repository is recommended as profiles are an input to the build important for reproducible (and performant!) builds. Storing alongside the source simplifies the build experience as there are no additional steps to get the profile beyond fetching the source.

I very much hope other languages/frameworks can do the same.

[0]: https://go.dev/doc/pgo#building

This is essentially how separation of duties(and concerns) looks like. And this is how some of the good examples of projects work. Specific techniques and tooling and specific boundaries of CI and CD vary depending on the nature of the end product but conceptually you are absolutely right.

A lot of projects and teams have been successfully using all tools mentioned in both CI and local dev envs. Portability and reusability of this “glue” stuff is a nice indicator of engineering maturity and culture within those projects and teams.

Low pro teams and individuals will happily abuse any new tool.

Mature teams stick to local-first CI - i.e. be able to run pipeline locally - and then translate to whatever is being used in their build and test infrastructure.

It resonates with me wrt muscle memory and ubiquity of “standard tools” that come pre-installed in majority of *nix distros including macos.

But there is a big BUT! Lately I have to use grep/find huge nested dirs and found rg to be an order of magnitude faster. Had to get myself comfortable with retraining the muscle memory. Worth the effort.

Some of these new shiny tools are meh for my taste. Delta for instance. Or helix the editor. But it is personal. Overall I love the competition. It seems like industry once full of innovators and tinkerers is lacking some shake up.

Have never been using facebook. Gave instagram a short stint few years ago to remove it eventually because of too high ads/friends ratio in the feed. No problem in organizing social life.

FWIW, whatsapp provides excellent capabilities (well, for 2025 where only old timers remember phpBB and can compare to it) for local community discussions and meetups, interest groups, parent groups, whatever group discussions and coordination.

While I agree with you I can’t imagine reforming the pronunciation of - say - “Ought to cough and hiccough through the ploughing drought”.

The man who got really pissed off by complete disconnection between spelling and pronunciation of words in English was George Bernard Shaw. Throughout his life he was an avid critic of the latin alphabet that was (and still is) used to represent spoken words.

He was so frustrated that shortly before his death in 1950 Shaw put the creation of a new alphabet in his will backed by grant in aid. And thus the Shavian alphabet was born. Enjoy:

- https://en.m.wikipedia.org/wiki/Shavian_alphabet

- https://youtu.be/D66LrlotvCA

I like weeks. Just 52 every year. Leap year agnostic.

I have a notebook for self reflection. Week-based. Since it is not a diary I call it septimary. Unlike a diary, the septimary gives me flexibility and every week I summarize something without mentioning specific dates or days unless this level of detail is really required by the context or significance of what needs to be written down.

Week 7 of year 2025 comes to an end.

I have finally found the value of llms in my daily work.

I never ask them anything that requires rigorous research and deep knowledge of subject matter.

But stuff like “create a script in python to do X and Y” or “how to do XY in bash” combined with “make it better” produces really good and working in 95% of the time results and saves my time a lot! No more googling for adhoc scripting. It is like having a junior dev by your side 24/7. Eager to pick up any task you throw at them, stupid and overconfident. Never self-reviewing himself. But “make it better” actually makes things better at least once.

Yeah. “Fast” in what sense? I tried printing all unicode characters in 80 symbol lines and scrolling the screen of several thousand lines does not freeze only in terminal.app.

Aside from some bizarre tests and benchmarks like this in normal life I don’t know what could be considered slow.

> there was not a risk

Yeah, buy a mistyped domain in question, setup recursive dns to build the picture of requests, build a “apigw” and route users’ requests to your own api gateway, continue until you phish users’ data or steal their money.

Mastercard was too lucky noone had done that and instead it was a good samaritan who secured the domain name to actually protect the giant corp and had reported it directly to them before disclosing it in public(as far as I understood the sequence of events).

And they are lucky there is zero impact(is it?) and unless this story goes viral outside IT/security research bubbles they won’t even care to correct their reputation and also help Bugcrowd find the definition of “ethical” and “professional” in the dictionary.

if you follow the sleigh morse path starting from the robin it will be 'united in protecting the kingdom'.

Did it get the 8 right? The linked article provides the wrong answer btw.

Loving this. I wonder how people even come up with an idea of truncating hashes. For what purpose or benefit?

What I find weird is searching IMG_XXXX directly on youtube returns you a number of videos with such title and half of the results are 10+ year old youtube SHORTS. There was no such thing back then. Just bloody videos! Does Youtube auto-convert short old videos to vertical shorts on users’ behalf?

Kubernetes is not a one size fits all solution but even the bullet points in the article raise a number of questions. I have been working with Kubernetes since 2016 and keep being pragmatic on tech stuff. Currently support 20+ clusters with a team of 5 people across 2 clouds plus on-prem. If Kubernetes is fine for this company/project/business use case/architecture we'll use it. Otherwise we'll consider whatever fits best for the specific target requirements.

Smelly points from the article:

- "147 false positive alerts" - alert and monitoring hygiene helps. Anything will have a low signal-to-noise ratio if not properly taken care of. Been there, done that.

- "$25,000/month just for control planes / 47 clusters across 3 cloud providers" - multiple questions here. Why so many clusters? Were they provider-managed(EKS, GKE, AKS, etc.) or self-managed? $500 per control plane per month is too much. Cost breakdown would be great.

- "23 emergency deployments / 4 major outages" - what was the nature of emergency and outages? Post mortem RCA summary? lessons learnt?..

- "40% of our nodes running Kubernetes components" - a potential indicator of a huge number of small worker nodes. Cluster autoscaler been used? descheduler been used? what were those components?

- "3x redundancy for high availability" - depends on your SLO, risk appetite and budget. it is fine to have 2x with 3 redundancy zones and stay lean on resource and budget usage, and it is not mandatory for *everything* to be highly available 24/7/365.

- "60% of DevOps time spent on maintenance" - https://sre.google/workbook/eliminating-toil/

- "30% increase in on-call incidents" - Postmortems, RCA, lessons learnt? on-call incidents do not increase just because of the specific tool or technology being used.

- "200+ YAML files for basic deployments" - There are multiple ways to organise and optimise configuration management. How was it done in the first place?

- "5 different monitoring tools / 3 separate logging solutions" - should be at most one for each case. 3 different cloud providers? So come up with a cloud-agnostic solution.

- "Constant version compatibility issues" - if due diligence is not properly done. Also, Kubernetes API is fairly stable(existing APIs preserve backwards compatibility) and predictable in terms of deprecating existing APIs.

That being said, glad to know the team has benefited from ditching Kubernetes. Just keep in mind that this "you don't need ${TECHNOLOGY_NAME} and here is why" is oftentimes an emotional generalisation of someone's particular experience and cannot be applied as the universal rule.

Let them fail. Does anyone today remember SCO chasing virtually everyone on Earth after unix licensing violation and copyright infringement claims? There is no shortage in KV stores with similar properties so if redisinc. wants to continue their path without delivering actual value to the customers especially to enterprise customers willing to spend money they know exactly where they will land in several years from now.

10 years ago I was using https://mosh.org/ to sutomatically reconnect ssh sessions.

I remember 2005 when I ordered the Ubuntu CD to the middle of nowhere with just a dial up connection and receiving a package from UK felt like magic. It was my first hands-on experience with GNU/Linux.

Time flies. Things change. "The year of Linux on the desktop" is always next year. I am grateful to Canonical and Mark Shuttleworth for all their work that was done back then. For a lot of people, Ubuntu became their entrypoint to the world of open source, tinkering and Linux ecosystem.

As someone who spent 6 years at the uni in the first half of 2000’s having access to the mixture of printed and digital materials, and as someone who has to read tons of different formal and informal docs on a daily basis, I can firmly say I don’t care about the results of this study and at least they are irrelevant to me. You need to get your tasks done, research and learn anything? Use whatever medium is available and fits your needs best and allows you to yield better results in a shorter timespan.

I won’t bother printing some programming language documentation and I miss the ability to search through text in printed books.

That being said, printed books smell good, and I enjoy the typographic excellence of them every time I grab a real book.

You know what. I just ran this query in gemini and it spit out “I'm not able to help with that, as I'm only a language model.” but just before that I got a glimpse of a real answer:

NFL Starting Quarterbacks for the 2024 Season Note: Quarterback situations can change throughout the season due to injuries, trades, or poor performance. AFC • Baltimore Ravens: Lamar Jackson Buffalo Bills: Josh Allen

But then it gets wiped and you cannot see it even in the drafts. The text above is from the screenshot I managed to make before the response vanished.

This non-deterministic unpredictable behavior blended with poor “safety” policies is one of those major “dealbreakers” that pushes me back from trusting any existing LLMs.

Brilliant. Reminds me of https://github.com/kelseyhightower/nocode but for general public.

Sometimes to get an idea of something you should simply stop thinking.

I have been using colima as a lightweight alternative to docker desktop and the likes of it for almost two years. Looking at the comparison provided on the orbstack website (https://docs.orbstack.dev/compare/colima) it seems to be not very accurate or at least requires some explanations/clarifications.

For instance: Low power/CPU usage is advertised as non-existent in colima. This is simply not true. Based on my perception I can't tell whether colima VM is running or not. Unlike docker desktop, especially with kubernetes on. Does not drain my battery, does not bog my CPU down unless I intentionally spin up something resource hungry.

ease of use/performance: not everyone needs GUI. colima is fine UX/devex wise with fast startup times. What does "fast network" even mean?

Linux machines/distros: not a fair comparison. colima stands for "containers on Lima" where lima is "linux machines" on macos. I.e. if you want arbitrary vms, use lima directly. colima is specifically built to spin up docker/containerd/k3s vms.

containers/kubernetes networking: this is opinionated and depends on a specific use case. In general I prefer the idea when my local kubernetes setup looks like the end production setup in the sense that I cannot mess up much with networking, access clusterip services directly from localhost because clusterip services are supposed to be accessible from inside the cluster itself, not from outside. loadbalancer IP is accessible through NodePorts anyways.

containers file access: there are plenty of ways you can access files in containers and images. But again, probably there are people who like to browse the guts of a kubernetes node in MacOS Finder. When it comes to files and networking I want to be able to re-use my toolbox used for dealing with remote kubernetes clusters and docker/containerd instances to my local ones. Creating a special case with convenient but non-standard ways to access files as if they were part of my host filesystem may be good for someone, but wrong for someone else because at times when something goes wrong this special case will work as an excuse for "works on my machine".

Please take the above as my personal experience. And I am in the herd of those who tend to keep everything as minimal and bare as possible with as much standartization/ lack of deviations across different environments as possible. Came to colima after years of minikube just because minikube's experience was no longer good with apple silicon. And there must be a very strong reason to switch to something new when what you have already is good enough.

Also, when it comes to GUI, what about Rancher Desktop?

Until the AI means “system thinking capability that can analyze the codebase and give real suggestions” I don’t buy it. Everything I have seen so far is waste of my time and resources and at best is useful for generating tests or docstrings.

I call it “PDCA over progressive JPEG” method.

The most fascinating thing about human intellect is how we all land in the same places, we just give them different names upon discovery.

In the top directory of my improvised library of electronic books, papers and pdf prints of valuable web pages there are three uncategorized files:

- How to read a book by Adler

- How to read a paper[0]

- How to study[1]

Serves as an entrypoint and gatekeeper. Reminds me every time how to spend my time effectively there.

[0] http://ccr.sigcomm.org/online/files/p83-keshavA.pdf

[1] https://cse.buffalo.edu/~rapaport/howtostudy.html

Nothing is new under the sun. Reminds me of the good ol' CoreOS (the original, not the fedora-flavored, although it also applies), Container-Optimized OS from GCP and the likes of the new wave with Talos and Bottlerocket.

My only wish is that immutable filesystem, read-only rootfs and most of the system with just a FEW exceptions, atomic upgrades/rollbacks would be more and more widely encouraged, discussed and adopted.

Just eliminating the worry of root partition running out of space is a life changer. From the ground up any "traditional" linux distro will force you sit and design how to partition the disk sooner or later. The guys above address this issue (and a huge bunch of others) for you.

Producing static PIE binaries is a bigger challenge still.

  $ go build -ldflags '-linkmode external -s -w -extldflags "--static-pie"' -buildmode=pie -tags 'osusergo,netgo,static_build' main.go

For anyone curious to delve into what is this and why: https://www.leviathansecurity.com/blog/aslr-protection-for-s...

I have turned my work iphone into a dumb phone. It has got only apps that cannot be removed, plus the apps that are provisioned through mdm. No icloud, nothing beyond what is needed for communication with my colleagues. Location services off, bluetootg is always off, wifi only connects to cirporate network.

My personal phone is less so dumb but I follow the YAGNI principle there as well.