Tootfinder

A panel. Is the EU over regulated? Can builders build here? Or will all the rules break industry?
The founders on stage start companies based from the start with compliance to regulation as a USP.
But regulators don't seem to understand the tech they regulate.
In survey 80% of companies chose to not enter the market because the laws are onerous. Government can't regulate industry that refuses to operate in their legal zones.
Companies may be spending 30% of their budget on legal compliance!
For my part I think good regulation is helpful but that much of the regulation here is bad, and some of it even malmotivated. Can encourage monopoly. The law is a mess.
#devWorld

Rowdy want to automate accessibility testing.
We all love accessibility, except those trying to enclose the web for private capital.
Wcag is the standard. Web content accessibility guidelines.
Playwrite is a node testing system and Axe-core is a Playwrite system to test accessibility.
He does a quick demo. Report seems nice. Axe seems like a good system for a11y checks.
#devWorld

The countries that aren't america need sovereign cloud infrastructure.
Can it be done with Lambda functions?
It can now, tcloud has cloud functions where its 'serverless' like aws lambda, which I've never used.
The key to serverless is of course that there are in fact servers. And starting them on demand automatically can be slow. Sounds like the EU tcloud has got faster at that lately.
He went through the web interface for using it. Lots of forms. Just like aws really.
#devWorld #soverignCloud

MCP stands for Model Context Protocol and is a api for agents. Storyblok has made their api into an MCP and Bernhard is here to tell us about that from the context of a psychologist.
Storyblok is a Content management system which can now have agents mess with it for you.
MCP standard connects agents to external systems. An mcp server runs to enable that.
Isn't the agent smart enough to use the api?
Well, they need documentation to figure out how anyway, so MCP rolls that in I guess? Avoid filling the context window with api docs.
Overlay specification from openapi let's you describe the api with a json doc.
The agent can query the server to search for commands, get the spec for the one they want, then call it.
"Skills" are troublesome because they gry loaded to dev machine and never updated. MCP stuff downloaded live on demand.
There exist tools to convert a rest apt to MCP but they apparently fail often because the machine ideally wants something different to a rest api.
#devWorld

Opening ceremony involves speaker asking the crowd individually why they are here. Mostly they seem to want to know how to use ai. And not lose their job.
Didn't ask me so didn't get to say that I'd like to learn how to defensively fight ai and prevent it spamming up the web.
#devWorld

In summary then, it is indeed quite like being at school. Half hour lessons on things that probably won't ever actually be useful to know in your particular job of varying levels of interest. Mostly pretty low interest honestly. Bumping into colleagues between lessons.
Learned the names of a couple of tools I might try. One google search would have gotten me those but I guess it's a question of thinking to look for them.
If you can judge the mood of an industry from a random selection of talks from a single conference then the industry is very optimistic that they can make AI write a lot of software.
It seems to think this is likely to mean fewer programmers rather than there being more software meaning more workers.
It wasn't as AI heavy as I thought when I first glanced at the program. Managed to mostly be not-ai I think.
Nobody talking about the ethical implications or suggesting joining a union and only one talk about the environment issue at all, it not really noting how much power the industry is about to take.
Liked having a few meals in amserdam with colleagues I never usually see (mostly remote workers, including me). The boss is pretty good at picking people really.
Get a day or so of holiday now too.
#devWorld

Hey one not on ai!
Continuous deployment stuff. Pushing to production.
Gitops is like devops only it pulls from git instead of being pushed.
Flux is a thing doing the pulling?
Its a kubernettes app. Install it on your cluster, and it'll watch your repo and build and deploy.
I think? I barely understand kubernettes, he said a lot if names of things I don't really know what they are. I've heard of Kustomize. Should look into that too if we ever get time.
He used k9s for the demo. Some sort of console ui for kubernettes? Dunno what that is really. Might be useful.
#devops #devWorld #kubernettes

React is a framework for web ui. There's a 'native' version which is supposed to let you do native apps as well as web apps.
Let's find out some pitfalls of that.
Component reuse is not always great, if it makes your component too complex. Means a bug suddenly can affect many things instead of one.
Performance in test vs prod can be diffent. Test data smaller than real data? Use shitty slow android test devices. Profile running code. Don't randomly change component ids.
Upgrage libs often. Gets more complex to do lots at once. Frequent cleaning means easy cleaning.
Write comments so people know who knows what functions are for and do. Knowledge leaving team as staff leave is bad.
#devworld

A man called Confidence thinks, presumably confidently, that the best interface for a ai agent is not a chat window on a website but... Email!
Chat is synchronous, app specific, dies with a tab close.
But agents tasks can be asynchronous.
Email has identity, a wake event when messages arrive, has a way to reply asynchronously, can attach files, and agents can email each other.
Email is already everywhere.
He has a js library to deal with email to agent messages.
He does a demo including buying a domain and setting up his lib to handle email from it. Interestingly, there's still a web chat to it. Ha.
All of which makes sense but. Email? Really? Surely something more secure and encrypted is better? Something with sender signing so random hackers cant email it? I thought this was surely going to be satire.
#devWorld

All the talks in sync at 30 minute intervals. This is too short really. They either have to dive in with jargon I don't get to explain in depth, or barely manage to summarize before the time runs out. No time for Q&A at all which leaves you wondering why not just watch it on youtube?
There's a hall full of stands from corporations of varying evilness giving out useless tat with trademarks on it. People go mad for the loot. Seems to make people quite excited to get a pair of socks or a fidget toy. I leave them to it. Don't like talking to salespeople even if you get free socks with trademarks on them. I don't wear logos in general anyway.
The AI bootsterism is strong, but not omnipresent. Plenty of talks on team management or deployment or progress in non-amazon cloud systems or whatever. Even if they can only be quite surface-detail and lacking depth due to shortness.
It's like being in school really. Flowing from lesson to lesson. Mostly being taught things that will be irrelevent or are boring or are unlikely to ever really come up.
Short break before street party now.
#devWorld

Saravanan talks to us about making greener tech.
Everyone else is burning fuel for ai like mad but Saravan wants to make things greener instead.
3% or so of greenhouse emissions are from IT. More than flying.
Cloud providers claim to be carbon neutral but this is changing with ai centers and often software written for them aren't used most of the time.
Green Software Foundation thinks software can be greener and offers courses and a profiler to tell you the carbon footprint of your code.
We can write for carbon efficiency. Electricity and hardware manufacturing included.
Power usage effectiveness can be calculated. How much is wasted on cooling vs compute, say.
Do your compute when solar is in excess instead of during times the power network is burning oil. Or do it in countries with greener grids.
Support older user hardware. Extend the life of end user devices.
Can't help feeling like these efforts are going to be drowned out by everyone else here burning tokens like there's no tomorrow. Which maybe will ensure there is indeed no tomorrow.
#devWorld #greenSoftware

Flower framework to deal with ai engineering work transition
Juliette is a game dev who was excited by ai then , oh, its disruptive and can software dev be ending?
So she made a model for thinking about the impact on teams and engineers.
It activities:
What is the work that might be affected? She has a list of things too long to read.
Define how each is done, standard best practice etc. How outputs of each are inputs to others.
Claude builds skills based on that.
Ai capabilities:
Can si do those things? Can it ever?
Probably not for many.
So what is the impact going to be?
If its all done by ai the job is thinking about the product not coding. That plus working out the odds its done it right. More importantly, philosophy! The job is to define what good software is and have the robot implement.
So sounds like she thinks were heading to the job being thinking about systems then wrangling robots.
But only if its reliable and cheap and not regulated away.
Which might not be the case.
Buy her book for the full list and analysis.
#devWorld #ai

Day two, and a much earlier start. Now Kevin Lewis from apify is here to tell us about how http 402, payment required, has been pointless till now but now maybe AI agents will pay when people couldn't.
Giving ai a wallet.
Dollar payments too slow and expensive for agents paying a penny for a page. Days for payment clearance. Kyc means every payer has to be human even.
X402 then instead. Stable coins on blockchain. Programmable , fast settlement. Pay in millionths of a dollar.
Or l402 to pay with bitcoin lightning.
Server sends bill and specifies payment details in 402 reply. Agent making request settles on chain then retries request with payment reciept.
Agents will want to evaluate if its worth the money within budget and etc. Give agents only limited wallet with disposable funds.
There's a mech for paying unpredictable prices. Over pay and refund whatever the excess .
He shows a demo. On how to pay a web scraper to scrape Instagram for you. For a dollar his service will give you 400 scraped profiles.
Paying for questionable services eh?
Not clear why an agent would pay another agent to scrape a website instead of just doing it itself.
No indication of how the agent can be prevented from leaking its private keys to injection attacks.
#devWorld

The man from Oracle says they are making databases more AI. Support for vectors in the db and systems to ensure secure access for agents.
He wants to encourage us to use one big (Oracle) db instead of lots of caches and nosql databases. The ai prefers thst, he reckons.
Oracle is like a smart phone. It can do everything. No need for separate devices for music and notes and whatnot. It can even do json now. And vectors.
Data privacy is often done at the app layer, which means a rogue ai could generate SQL to bypas that. So do access control in the db direftly. Which is probably better even without ai honestly. This 'deep security' might work well.
Ai is happier , according to the man from Oracle, if it just has one Oracle db without competitors software at all.
Fancy that.
#devWorld