As much as I like OpenClaw, the terrible UX/DX makes it pretty frustrating to work with.
Also, why does it feel like Minimax is so much worse than Claude? The benchmarks make it seem like they’re at least comparable in performance. Is it the training, or a harness, or my personal bias?
You are absolutely right, and to be absolutely honest: I have no idea on how to deal with fatigue of the reviewer.
My current idea is that in practice you will have to combine as many safety systems as you can and only adding human approval where absolutely critical.
Excellent question.
Hope is not a security strategy.
New post on building safe AI agent pipelines -- Plan-then-Execute, human approval gates, the Lethal Trifecta.
www.amathia.org/blog/safe-ag...
#AIAgents #Security #EUAIAct
New in BreakLeg: deterministic process nodes, typed data flow labels, memory subtypes & decision annotations on edges.
Security Analysis and System Design for Agentic Pipelines, all in one app!
Try it: trifecta.amathia.org
#AIAgents #Security #OpenClaw #SystemDesign
New in BreakLeg: share your agentic pipeline designs via URL: Just click Share and send the link. No server, no signup, everything stays in the browser.
Try it: trifecta.amathia.org
#AIAgents #Security #OpenClaw
Scared that your agentic setup might leak all your secrets? Yeah me too. So I built a tool that visualizes the dangers.
Not scared? Check out my tool to see why you should be.
Drag, drop, instant feedback. Free, runs in your browser.
www.amathia.org/blog/breakle...
#AIAgents #Security #OpenClaw
Was scared of the not at all present security of OpenClaw. Still am. But I spent some time trying to make it less scary.
Read here www.amathia.org/blog/sandbox...
📢 React Vienna Meetup in April:
🗓️ April 8rd, 18:30 CEST
📍Eversports, Jakov-Lind-Straße 13, 5th Floor, 1020 · Vienna
💬 React.use(Compiler) - Can we finally forget about dependency arrays? by Max Zauner
💬 Building a multi-domain SSO with FusionAuth and NextJs: Challenges and learnings by @cezar.is
Yeah, for a few months now. Liked copilot, but Cursor‘s take on it seems better to me, especially for repetitive tasks. If you like copilot I’d recommend giving it a shot. I came from VSCode though, switching from another IDE might be more of an issue.
Very neat, love it!
Vue 3 is fine. Nuxt 3 is harder, GPT for example likes to give Nuxt 2 code. Overall it’s fine though. It’s great for laravel as well.
I recently had some issues with hyphenation which I then just gave up on in Firefox.
Just out of curiosity, what glyphs are you having trouble with?
Keep it to avoid confusion.
Exploring ways to make nice looking packery-like image galleries in WordPress today. Loads of plugins that seem cool but I’d like to make one without paying for it. This is not a complicated enough feature to warrant the expense, I feel like. Let’s see if I’m right😅
Today was my first encounter with a 360 deg meeting camera. It saw me while not even looking at me! Spooky but impressive.
Got it working. It was (unsurprisingly) my mistake. Still had some requires in a few places, which Vitest apparently does not condone at all😂
Tests now run definitely faster than before: 10s instead of ~17s before with Jest. But: after the tests it says 22.7s, which definitely was not the case.
Still nibbling away at the Jest to Vitest migration (while knowing little about Vitest). Making progress, but am now stumped by module not found errors. Seems to be a common issue though.
Alright, I understand now. Thank you for your insight!
Sorry about that. I’m just wondering which cases there are where laravel is the best tool for the job while not doing any content management. Genuine question, I think I’m just not imaginative enough
How did laravel come into play here, if not for content management?
Very interesting! What kind of stuff are you building if I might ask?
Impressive that you manage to do this every day. Incredible consistency!
Very cool! How much time do spend on average per tool?
Comparing yourself to others is a quick and certain way to unhappiness.
You compare your whole being to just a fraction of the being of the other person, which massively skews the results in the others persons favour.
NOT comparing yourself to others might be one of the most important but also one of the most difficult things to learn.
The reality is that most of the time you know very little about a person, their history or what’s going on in their heads and lives when nobody is looking.
Today I began migrating to Vitest from Jest, mostly because I heard good things about Vitest. It is not going smoothly at all. Wish me luck.
Currently resisting the desire to rewrite a small Vue page of mine with HTMX😅 I know little about HTMX but I think it’d be a good fit and I need a reason to learn it.
Fixed a bug today. Parsing a text file I used for..in to iterate over an object. Forgot that the property name is a string, even if the property is a number, and treated it as a number, leading to a multi hour bug hunt. The second time that has happened to me. Bad practice to use numbers as keys?
I can’t find a stackexchange policy like that mentioned anywhere do you maybe have a link? If they have done so information about the framework can still be found on the framework documentation page e.g. ChatGPT can also look up more recent information if instructed, but the quality varies.
Already happening right now that e.g. GPT suggests code that only works in older versions of certain frameworks. Maybe LLM maintainers will be able to feed it fresh data more quickly? Or LLMs will draw on information from the web in a more capable way? It is an interesting question.