snowfield

Developers will always hate to ask the system owner for changes to permissions don’t think you can just fix that.

If the company has enough grc red tape, integrating with slack can become almost impossible I can imagine

Yes, in a world of dynamic virtual teams and cross cooperation across teams. «Teams» is an ancient construct

Just vibe code it bros

Yeah that’s what they try to do with the latest coding agents sub agents which only have the context they need etc. but atm it’s too much work to manage contexts at that level

A developer can blast millions of tokens in minutes. When you have a context size of 250k that’s just 4 queries. But with tool usage and subsequent calls etc it can easily just do many millions in one request

But if you just ask a question or something it’ll take a while to spend a million tokens…

Because your national deficit is 70% of your gdp you kinda have to

Grok can ingest images and do analysis in real time. But yeah it's going to be real hard for it to relate that image to a current event.

Yeah, mail is the primary source of this.

Once communication with my customers moved to teams. I've had a very hard time to find historical agreements and decisions.

I try very hard to create a robust system for ADR logging now. And not just for system architecture. But for all decisions and agreements in my projects and across changes.

Don't think that really works.think that's been debunked

"I use Remote Desktop so everything is accessible from every device"

But it's not, it's a database. That is annoyingøy hard to move around and version control

I'd recommend chrome mask instead. It's written by a Firefox engineer

https://addons.mozilla.org/en-CA/firefox/addon/chrome-mask/

I think the argument is that clippy would totally have done that if it was an option back then.

Agreed, I know Networks, requests, protocols, Auth flow etc but is been years since I actually coded stuff

This is magical to me.

I love cursor, I use it to deploy docker packages and fix npm issues etc too :p

I use some guardrails, like SonarQube as static code analyzer and of course some default linters. Checks and balances

I often find myself clearing the context when dealing with llms to get a fresh take. Often it just has so much context reinforcing its previous decisions.

Not sure if the author tried to just start a new thread. But anyway, for now you always need to keep an eye on these things and manage it if it follows red herrings or ends up in some logical loop

Sidenote : newlines is one thing tat can be quite tricky for llms in general.

This article is ment for companies who want to save a buck on aws costs.

Played until 27. All the levels were pretty easy, 15-30 seconds per level. Mainly because of the "par" feature

You can host authentik with one click in docker. It's super easy to set up

This feels hopelessly dated and given the reception it doesn't seem like it works that well

Isn't this a standard feature these days?

I use it quite often. But just small stuff like simple automations, userscripts, docker config that sort of thing

Love d2, using it for years. I use it whenever someone asks me to create a visualisation. And I regularly tell people about d2 trying to push it internally haha

Enterprise is way too expensive though. And I can't natively render it anywhere. So I'm limited to only personal use.

3000usd / year for a single TALA license is... Well, not justifiable. And I'm not sure how much d2 studio is on top...

I'd love to test TALA though, but even the personal license at 120usd is steep for drawing some stuff.

Just giving a helm chart or system description to an llm and having it one shot a perfect diagram which code I can easily view and edit is also a big gamechanger for me

Sucks to not have it just render natively anywhere, I think confluence cloud can do it. But I think the add on is a paid addon.

D2 can be locally compiled. Llms are pretty good at generating d2 too. So all you need is to tell the LLM what to do tbh.

Not my project but no need for an mcp I'd think

I'd assume that behind the scenes the models generate several passes and only show the user the best one, that would be smart, as to to make it seem their model is better than others

Is also pretty obvious that the models have some built in prompt system rules that makes the final output a certain style. They seem very consistent

It also looks like 40 has the temperature turned way down, to ensure max adherence, while midjourney etc seem to have higher temperature.more interesting end results, flourishing, complex Materials and backgrounds

Also what's with 4o's sepia tones. Post editing in the gen workflows?

I don't believe any of these just generate the image though, there's likely several steps in each workflows to present the final images outputted to the user in the absolute best light.

High context is pretty normal these days though, as you keep interfacing with the llms the context window just grows. And with mcps and RAG is trivial to get 30k contexts++ in every query

This is a very interesting concept. I have a few questions though.

Even if the agent has mcp tools avaible, you can't guarantee that it would necessarily use them right? I imagine that you'd have a lot of new tools that do nearly the same things.

Also how does this scale? If the agent has like 300 tools.

Mail is great, but messaging has its purpose.

I use ai to write diagrams too. But I use d2 instead of mermaid

Really cool, all my experiments with local speech has been spotty. Keen to try this out.

Can it run on a gpu? Would it be faster?

AI training doesn't work like that. you don't train it on context, you train it on recognition and patterns.

I set up an MCP for jira, to hopefully avoid me from ever opening jira again.

i will just talk to the LLm, and ask it to set a task to in progress, tell me what to work on next, add tasks to sprints, tag tickets with labels / link them to epics, search for issues, create bugs / tasks etc

seems mildly helpful, with some slack mcp etc it could be useful for teams i think

i really don't want to just buy an agentic workflow from a website, so its important that i can just replace the model with whatever and that i have control of the mcp server