15:[[48771515,{"by":"yehiaabdelm","descendants":206,"id":48771515,"kids":[48779784,48780140,48779017,48782072,48771907,48779404,48780818,48771922,48774103,48772037,48772023,48779308,48772274,48780664,48771844,48771558,48771908,48780254,48772049,48771846,48772031,48773615,48780323,48781568,48772067,48779870,48774635,48813788,48782539,48779198,48772655,48772050,48779159,48779429,48780572,48771813,48865107,48801622,48779519,48780725,48776921,48780615,48772171,48780665,48771940,48779774,48780635,48780512,48781102,48774147,48773463,48861468,48773801,48779629,48772253,48779307,48780148,48780663,48781607,48772551,48814494,48779953,48772025,48779985,48781823,48780687,48771836,48771910,48779833,48771969,48773778,48840418,48779215,48780877,48781699,48772203,48771683,48804058,48781629,48790124,48772113,48771785,48772201,48786954,48772226,48782641,48771738,48790263,48779375,48785505,48772149,48780190,48780621,48779767,48779892,48772247,48786289,48771982,48773409,48772574,48831094,48806894,48782296,48780544,48785971,48794004,48782408,48772013,48771644,48772527,48800273,48772632,48788537,48792568,48791893,48771945,48812754,48793079,48799908,48780848,48787253,48787133,48781470,48862733,48784499,48780079,48780259,48772301],"score":212,"text":"I'm a bit annoyed by the feeling that we're kind of stuck when it comes to using LLMs for programming.

I use Claude Code and Codex, but I haven't been able to enter flow state like I can when I hand write code.

This is kind of ironic to me since AI should be a bicycle for the mind, but right now it feels like a bicycle that just brakes abruptly every couple minutes. I stop, wait, review, prompt again.

Is there anyone exploring something fundamentally different than the prompt response loop we have today?

I actually think the idea of a tab model is directionally better than prompt response.

Would love to hear about any startups, personal experiments, etc.","time":1783059704,"title":"Ask HN: Is anyone experimenting with different ways of using LLMs for coding?","type":"story"}],[48021462,{"by":"shahisoft","descendants":0,"id":48021462,"kids":[48033464],"score":2,"text":"With the rise of SearchGPT, Perplexity, and Gemini, the goal of content is shifting from "ranking on page 1" to "being cited in the answer block."

I’ve been working on a tool (https://aibg-intelliagent.com/) that uses a private RAG (Retrieval-Augmented Generation) system to pull from local site data/PDFs to ensure the AI output is factually grounded enough to be "citeable."

To those working in this space: Are you finding that structured data (JSON-LD) and RAG-based content are actually moving the needle for AI-driven traffic, or is traditional semantic HTML still king?","time":1777983478,"title":"Ask HN: Are you optimizing content for AI Search (GEO) vs. traditional","type":"story"}],[48018109,{"by":"indieDevAttempt","descendants":0,"id":48018109,"kids":[48107428,48019247],"score":7,"text":"$16","time":1777956087,"title":"Ask HN:How's you experience building an app on Cloudflare? (Workers, D1,R2 etc.)","type":"story"}],[47982934,{"by":"hysan","descendants":2,"id":47982934,"kids":[47983006,47982954,47983806],"score":2,"text":"I got two emails from Google recently. One saying, “You’re now using Gemini on web” and the other saying, “Welcome to Gemini”

I never signed up for Gemini and I see nothing in those emails about how to disable or remove whatever it is that I supposedly signed up for. I did find others who have gotten the same thing out of nowhere and who have helpfully pointed me to a settings page to delete and turn off Gemini activity. However, I had done that and gotten that second email. Does anyone know why I’m getting these emails, how to unsubscribe to them, and how to get rid of whatever it is that I supposedly signed up for?

Other info - I don’t use Android. I don’t use Google as my primary search. I do use it as a fallback which is rare. All I really use Google for is Gmail and YouTube.","time":1777691319,"title":"Ask HN: Forced into Gemini on Google Account?","type":"story"}],[47536498,{"by":"herol3oy","descendants":0,"id":47536498,"kids":[47687028],"score":2,"time":1774563167,"title":"Astro 6.1","type":"story","url":"https://astro.build/blog/astro-610/"}],[47636026,{"by":"LordKode","descendants":0,"id":47636026,"kids":[47642695,47671731],"score":4,"text":"I was auditing a React site recently and noticed that the 'Wall of Love' widget added 400kb of JS and 3 iFrames, tanking the LCP. It felt wrong. We spend weeks optimizing every pixel, then we throw it all away for a slow, black-box testimonial slider.

I decided to build Reviewskits [0] to see if we could move to a purely headless, data-first approach using Bun and Hono.

The goal: Fetch raw JSON, build your own UI components, and keep a 100/100 Lighthouse score.

My question to HN: Do you prioritize the 'ease of use' of pre-made widgets over performance, or is the industry ready for a developer-first, headless infrastructure for social proof? I'm curious to hear how you handle this in your projects.

[0] https://github.com/reviews-kits-team/reviews-kits\n[1] https://reviewskits.com/","time":1775279951,"title":"Ask HN: Why still embed heavy 3rd-party iFrames for simple social proof?","type":"story"}],[47656335,{"by":"jer0me","descendants":61,"id":47656335,"kids":[47659598,47658579,47660276,47658764,47661498,47659081,47658381,47659787,47658385,47685279,47659148,47671475,47684685,47658318,47660819,47660556,47659267,47658803,47658208,47656517,47658906,47660026,47656475,47659908],"score":63,"time":1775442801,"title":"Does coding with LLMs mean more microservices?","type":"story","url":"https://ben.page/microservices"}],[47656501,{"by":"keepamovin","descendants":192,"id":47656501,"kids":[47657126,47657301,47657132,47659070,47658734,47658911,47658350,47656781,47658498,47657948,47670754,47659836,47656646,47657097,47659455,47659938,47659327,47662086,47659625,47658081,47659819,47658226,47659857,47659793,47661165,47670416,47657321,47658986,47659221,47659873,47657849,47660926,47658192,47663841,47663747,47669818,47659087,47660551,47663534,47657573,47659534,47661629,47659281,47673703,47661780,47666925,47656898,47658377,47660121,47666660,47671965,47659450,47664918,47657136,47661537,47663932,47658777,47657232,47657940,47660338],"score":309,"time":1775444602,"title":"The 1987 game “The Last Ninja” was 40 kilobytes","type":"story","url":"https://twitter.com/exQUIZitely/status/2040777977521398151"}]]

user:	reedycat
created:	February 26, 2026
karma:	1
about:	Web developer focused on Astro and Cloudflare.