I'm surprised how freely I speak with genAI. I would have been so grateful as a young, dumb dev to ask an LLM "I'm not understanding what this is, can you help me or guide me to ask my seniors the right questions that make me NOT sound like an idiot"
As always: it's not the fact the LLM is used to help develop a product. It's the lack of building upon that. That thing already exists, so the LLM could spit it out, do something NEW or why bother
Reminds me of that boring corporate art trend
I genuinely distrust anything that reads as AI. I use LLMs every day, and if I see a webpage riddled in badges and highlight cards, I assume very little consideration was made for the final product.
Idk why people feel alright publishing something completely written by a bot
Sure it sounds good... until you see the writing style replicated all over. I've been perusing insta a little more lately and it's relentless. The wall of text on a post feels so fraudulent.
I expect authenticity will soon become the appeal for online content... Maybe just hope.
I'm not even bumping hard against the budget, like I know others are exceeding quickly, but it's impossible to track WHY.
How is my token usage $200 and someone else is $1000?? Why penny pinch at these costs? I'm not gonna suddenly 4x my usage just to do it, but if I need it...
I've become highly aware of my LLM "usage" cost since the org has licensed us and this has weird pressure for me.
If I'm using a lot, how does that reflect on me for offloading work to this tool?
Will it be expected that my output is higher if I use the agents heavily?
We've set a standard already for increased output, and if devs are going to be budgeted, the company will force their hand to use personal accounts to maintain
Consider the token burn issues plaguing anthropic. devs have so little control besides swapping to haiku
the way Netflix, Shopify, Stripe, and Azure are handling all the facets of the problem I'm trying to solve, plus a ton of medium articles and docs around the tech. 455 sources in 6 mins while I rode to the store to get dinner stuff
None of it was on the first 3 pages of Bing
I'm so frustrated with tech research now that I am forced to make an LLM do it for me. I can't find how companies are strategizing a tech stack because bing exclusively surfaces adverts for products that do it for me. I just asked Claude the same thing and it found & summarized
The intro is a tad bait-y, but there's a SOLID 35 mins of great perspective and guidance in here, and I want to emphasize how temporary this is. Like he says, prompting like you did in January is not going to be the same going into March, and it will further change by Q3
This breakdown is incredibly valuable on building a skill of your agentic engineering, and I suggest practicing going through these steps on a new project, then the next one, then the next one
https://t.co/Qi4Burudp8
Using Amazon wishlists? This thread is for you!
A recent email to users stated that shipping addresses will start being revealed to 3rd-party sellers starting March 25.
Currently sellers can only see information on recipients' cities and states...
1/4
considered
Anyway...
Now my ideas don't get the chance for me to mull them over because I can already have Claude or Codex build it out before I doubt myself
PROTOTYPE EVERYTHING. Every version. Build it all, you work out the best one in less time than it once took to define it
I've always struggled knowing if an idea is a good one. Confusingly, I've always had a big ego, so I don't know why I would think my idea isn't impactful enough.
Turns out, once I mull over an idea enough, I land at the conclusion it's such an obvious idea that it's already been
When I entered the field, all I wanted to do was unwrap what my coworkers called "black boxes." Anything from legacy products on our platform to the frameworks we used to build new products. Today, I feel I'm watching the black boxes form around LLM Agents
I am fascinated to watch it form, and I'm trying to absorb every bit of it. I don't feel like an intern again, because I understand much more and can better judge what exactly I need to learn to keep up. Ultimately, this is could be my doom. Keeping up won't cut it
I had the advantage then, when I felt that to understand anything I had to know everything. The way school (sorta) taught me to think. I gotta get back to that...
The upcoming hurdle is abstracting the agent flow. How does someone with a problem (the prompt) get across the solution (acceptance criteria) on a website builder?
Today's models do pretty well with assumptions, as their discernment is likely based on the statistically common
coworkers were chatting about building a site for someone that otherwise would be relegated to contracting it out or using a website builder. Successful agentic building is still primarily in the hands of tech-literate people. A website builder with a prompt box doesn't suffice