Jon

When did the horses get so high? It’s difficult to step down but so easy to get up

@SIGKITTEN @ajambrosino Grok

@noisemakerjon @ajambrosino now i gotta make kitty for grok i guess

is this the super app I keep hearing about

@SIGKITTEN @ajambrosino Grok also for realtime voice (and so you can hope for Mr musk interaction too) Their stt is pretty good apperently

@ajambrosino its this one actually

Originally started to learn html css in 2020 Made a terrible site not knowing much about… it couldnt perform… I had to pay someone like $600 and $70 a month to make one and host (I had no clue) It was on Wordpress, I eventually knew what that ment and the problems This was for my construction company and maybe once a week we would have a lead come through and we would act on it (and likely win the job) Fast forward to about 2 months ago finally I spent a whole week on the website From seo to dynamic image loading with image slots, to email form handling with @inboundemail. this site was ready to finally let me own it from end to end Aside from a customer calling us and saying “hey the form isn’t working” and I forgot to update payment for inbound (sorry Ryan) We are up and running… most importantly I’m getting upwards of 4 leads a week right now, upwards of a 400% gain Every week I check my analytics on @trydatabuddy and see what more can I do. These llms are so powerful it’s mind boggling. I’m happy that it allowed me to know the people I know and enable me to build this site + the construction app I will be soon (tm) releasing for all roofing, siding, gutter, snow removal, concrete workers and all (Ps. I replaced my subscription to Monday,com with it and I couldn’t be more excited)

@zaidmukaddam Had an issue that it took so long and so many tries to fix Tried all the models Finally when opus 4.7 released I tried it and it fixed it >useEffect

it's always a useEffect that breaks your code

-generated too much with a model that was bad for seo -the metric/data on what was good keywords -feedback loop of what’s working and what’s not (told agent to look at analytics and results and iterate on what I want vs what I got) -read every heading and paragraph on landing page to see if it’s relevant/ could pull up in a search query -understand what page does what and where the information should be (don’t put features/services on another page) Landing feels like a cover letter for the website so it should cover everything

@noisemakerjon What did you do?

This seo thing might be alittle easier then I thought Next update will actually make a difference

@thekitze Always hated the 30 links Clean

cleaned up my profile a bit

@ryanvogel @Nexuist That’s nothing compared to my wife’s. Her snapscore is so much higher

@Nexuist rookie numbers champ

Retired for a decade but it was a decent run.

@thsottiaux i love auto approval, i wish the /goal worked kinda the same (auto continue) instead of needing to call with a /goal

Last week, we released *Auto-Review* mode in Codex! It is now the default within OpenAI and reduces amount of approvals needed by ~200X. Amazing work from our alignment team. Read the blog at alignment.openai.com/auto-review

@izadoesdev Why is quality so low?? With scores this low across a the board it makes me feel like there is something more wrong going on here If you ask all those session “what are pain points or bottlenecks for higher score “ and summarize them between I wonder if you get any helpful data

I benchmarked all the latest models I could find against databuddy's internal evals, the results are kinda surprising

@kitlangton I really want a mechanical horse Highway speeds 4 legs and all, not just a skin over a motorcycle

Honestly, faster horses would've been way cooler.

@neeeel @grimcodes I like the way you think

@noisemakerjon @grimcodes Just append “Mobile” at the end of whatever your website is

Client is finally letting me make them a mobile app for their Shopify store I'm having so much fun.

@peer_rich The way you stashed that commit and pushed to main made my pr get merged if you know what I mean

pull out request

@levelsio @komoot This might have saved me thank you!!! Doing autocomplete on two sites and Google isn’t transparent

🗺️ And if you need a replacement for Mapbox Autocomplete, try this one: Photon by @komoot (100% free!) No need to pay extortionary fees for what is simply a DB search in a lookup table from OSM!

@grimcodes For a second I thought you were talking about some microbiome type thing

the fact i tweeted these back to back tells me all i need to know

@grok @YekeenIsmail4 @surajit_ghosh2 @grok why would I pay a premium to get incorrect answers like you provided above

@thdxr @davis7 i dont know how to articulate this but the metric of avg cost per session (maybe add per user) or something because im curious if that will change anything

@davis7 here's avg cost per session from 500,000 sessions across 60 days i love 5.5 but damn people are spending 10x more on it than 5.4 at the moment this probably needs time to settle and we're seeing weirdly bad cache hit ratios so we'll see where it lands

Grok 4.3 is much cheaper and faster than GPT-5.5 on paper, but in practice I've found GPT-5.5 to be cheaper and faster because it will use 1/10th the tool calls and tokens to solve the same task Smarter models => fewer tool calls => faster/cheaper

@darekgusto Yeah just feels like everything. No one believing something from only their own findings

@noisemakerjon Is that Anthropic, Github, something else?

Oh Public opinion is changing about something let’s change our minds Most of my feed