James Dawson

Built a radial menu button to try it out. Inspired by @designloomco design.

@mcuban Knowing when it is wrong, when to call out its over confidence and when not to trust it's facts. It is a hard skill.

I’m coming to the conclusion that the biggest challenge for Enterprise AI, and AI in general , as of now, is that it’s still impossible to make sure that everyone gets the same answer to the same question, every time. Which is a great response to the doomers. AI doesn’t know the consequences of its output. Judgement and the ability to challenge AI output is becoming increasingly necessary, and valuable. Which makes domain knowledge more valuable by the second. Am I wrong ?

@AndyGoode10 Andy I am a big fan of you on the rugby pod, but this is a poor response. You are a grown man, not a 17 year old. That he got this lowshot reaction from you shows he hit a nerve.

Some of us have played the game and some of us haven’t, imagine being a Bath fan seeing the lack of replays and for blatant yellow cards. My job is to say what I see, your job is to sit in your Mum’s basement and watch

@unk_data @yodontcalm Mate, the AI will have its own weather balloon agent deployed on the edge.

@yodontcalm "What's the weather in Paris?" AI has to call some service or talk to AI running on some service, either way they may name it different things like MCPs but it's just an API.

Sometimes I think even Naval doesn't know what he is talking about. What do you mean AI will replace APIs?

In design, when a thing is everywhere we perceive it as cheap. AI generated landing pages are everywhere.

@MeghanBobrowsky Jensen Huang also saying "AI is not going to replace people"

This sounded familiar to me so I looked up Mark Zuckerberg’s comments from Wednesday’s earnings call: “My view of AI is very different from many others in the industry. I hear a lot of people out there talk about how AI is going to replace people...” (1/2)

For everyone who is worried about losing their job, this is google right now.

yeah, it a force multiplier to people with 0 experience and people with 10 years experience. Experienced software engineers will build it better, faster, more performant, more reliable, more secure and easier to maintain... because they know how. Even with AI, people with no experience will have to figure that out. And to the people who say "Fix all security issues" will work... I found a critical security issue on a $20 million AI startups production app last week... So, it is still happening and I would say the bugs and security issues are a force multiplier with AI as well.

@zeeg Agreed there’s still a massive gap between generating code and architecting real production systems My point is just that the barrier to becoming a builder has collapsed. Pre-LLMs, I realistically never would’ve built and shipped products while working a non-engineering job

imagine not having expertise in software and trusting (literally) anything these agents output

Being a good software engineer today is having the knowledge to call AI out on its subtle bullshit decisions.

Who is starting out trying to learn to code? I would love to help you

@mattpocockuk /lament Matt Let's see what claude comes up with

I feel sorry for Claude Code I know they're not the one. I'm not overcommitting - not investing too hard I wonder if they know I'm pulling away

Is it embarrassing to of never seen the confetti when you get all 100 in lighthouse?

What if AI is 10x better at exploiting/hacking code than writing/fixing it.

@joshpuckett Couldn't agree more, but I am really feeling the pressure to churn stuff out with AI. As soon as I am employed again (got laid off last month), I will take this course!

A confession: I've spent most of my nights and weekends this year so far on a bit of mission. I wanted to try and show that if you apply an uncommon level of effort and care to what you make—really pour yourself into it—that it just might resonate and stand out amongst all the noise. It's too easy to succumb to the pressure of moving faster, automating more, parallelizing, having your agents do it for you, or whatever else the latest thing is. But I see AI as just one more tool available to craftsmen and builders who take the time to make necessary, useful, and beautiful things. And like any tool, it's one that should be used with intention and consideration. When I wrote the below for InterfaceCraft.dev back in January, I really meant it. And I believe now, more than ever, that a better way is possible.

@wagslane Just add a woke/not woke toggle for them to select. Claude could 1 shot it easy

I literally can't win with pronouns on Boot dev, it's insane. We have people cancel over NOT using "they" in lesson text and people cancelling over us USING "they". > In one of your lessons you refer to a single person named "Logan" using the third person plural pronoun. > > At this point without assurances that the developers intend to correct the course material to correctly use the grammatically correct generic masculine as the correct generic pronoun, I have no interest in financially supporting the intentional cultural vandalism of my personal linguistic heritage Here's your refund, you won't be missed. I DON'T CARE about this. I don't care. WHY DO PEOPLE CARE. These are not real people in the lesson text. They're genderless EXAMPLES. Sometimes we say he/she, sometimes we say they, it depends on context and which maintainer happened to update it last. wwhwhhwyyyyy dooooo weee careeee sooo much. get a life.

Soft Navigation LCP shouldn't be scored like hard navigation LCP. Once the app shell is downloaded, changing view should have a much higher speed expectation: ⚡ 0–200ms: Excellent 🟩 200–500ms: Good 🟨 500ms–1s: Needs work 🟥 1s+: Slow

@rauchg Hey Guillermo, a company (I believe) you are invested in has a serious XSS vulnerability. I have DM'd you with the full details. They are ignoring me and this must be fixed.

I just injected AND executed a JS script in a $16 million dollar AI companies dashboard. The worst part is, they give me a link to share the injection with the world. Absolutely crazy. I will share it here, once they fix it...

Getting AI to consistently produce high code quality is the bottleneck. I keep finding subtly shit code.

Better AI prompting in 4 steps: Q&A first: let the AI ask everything it needs Agree on a plan: align before a single line is written Feed context gradually: drop relevant docs in at each phase of the plan. Execute. "No mistakes" is stupid.

Your coding process should be as deterministic as possible. How many of your agents /skills could be scripts or handled with linting? A lot.