Erwin

Tailscan 2 is live and on Product Hunt today 🚀🔥! This update contains lots of new features and an improved UI, designed to give your workflow a serious boost🙌! Read on here 👉 producthunt.com/posts/tailscan… PS: there's a limited amount of launch discounts available but be fast!

@kylegawley Friend of mine @lucaslovexoxo

@Erwin_AI hadn't heard of it but looks good, who made it?

Just purchased a new domain 🎉 No SaaS this time Building my first native MacOS app

@PascalPixel yes, deactivated itself

Has anyone actually gotten a local model to do anything useful at all? If so how?

@thorwebdev @elliscrosby @OpenAI @sama @zhennydez My use case isn't just speed but mostly data enrichment like writing summaries from large portions of very complicated text. Gpt nano models don't quite cut it, and thinking models are too expensive as we process ~1b tokens a month.

@Erwin_AI @elliscrosby @OpenAI @sama Would love to learn more about your use case! And also wonder if Gemma 4 31b on cerebras could be a fit for your: cerebras.ai/blog/gemma-4-o… Maybe @zhennydez can hook you up with some credits to test!

I've said it before but will say it again: We need a new AI model that is in the same league as gpt-4.1-mini. Sharp pricing, end-to-end and not just good for data extraction (like gpt-5.4-nano), and option to disable it (so not gpt-5-mini). gpt-5.6-nova? @OpenAI @sama

@elliscrosby @OpenAI @sama Pretty remarkable that Gemini is owning this entire scoreboard (@thorwebdev) It's not exactly, but definitely aligned to my main pain point. Cost is obv the other aspect.

The current selection of models that are genuinely fast + smart is really unsaturated right now. It's pretty much just Gemini releasing models for that spec. I made this benchmark (LLMs playing bullet+blitz chess) to see which models would work best at agent routing - I pretty much built it in anticipation of GPT 5.6 Luna but it ended up being way too slow... The top 15 is various Gemini Flash/Flash Lite models + Qwen 3.5 Flash 😅 springprompt.com/evals/bullet-c…

@OpenAI @sama @romainhuet @danizeres oh! Do either of you know who could be of help here?

@OpenAI @sama Repeat of x.com/Erwin_AI/statu…

@ImSh4yy Man.. At this point I feel bad for the people at Google. Not cause they can't deliver, but because they keep trying and pushing and keep getting one-upped before they have even released their model, then they release anyway and get rekt on X every time 🥲

Gemini 3.6 Flash: More expensive and worse than GPT-5.6 Luna.

@jonathan_wilke @supastarter I've jumped ship with Tailscan over a year ago. It's obliterating the template/devtool space. Best to start something else ASAP in a different industry.

Always trying to share the full picture, not just the wins. A few months ago, @supastarter was doing ~$13k/month. The last 30 days just ~$6k. Not going to lie, it's starting to worry me. If you were in my shoes, would you double down on marketing or pivot?

@Annatngm Yeah, same thought. It's tiring. x.com/i/status/20743…

Every guy wants to show me his AI setup and explain how much time it’s saving him Unless you’re training foundation models, chances are your personal AI workflow (or your pre-revenue startup concept) isn’t all that unique Congrats on discovering prompts & Zapier 🙄

@schuchmann_r Unmotivated, lazy people will remain unmotivated and lazy.

I've been consulting for quite a bunch of big corporations. Now what I noticed every time, productivity there is non-existent. It's quite nice, you can output 0 and it's normal (I mean for them, I don't like that mindset). I was walking in Sydney this morning where there are big offices of Salesforce, EY, etc. In my mind it's like, many people in the indie-hacking, vibe coding movement can easily replace a big percentage of employees in these companies. Not to scare anyone, but I am a strong believer that all these peoples in their time off need to start building something, just to get a feel of it. Better to start now and be ahead of the other 90% of population than fall behind and risk your job.

First impression after using for a few hours: It's a bit slow at times, but really really good. Really. Good.

@thekitze Dude...

vercel api doesnt expose the env vars so now i lost env vars for 20+ projects without a backup 🙂 meaning all the purchase flows etc and everything else is broken 🙂 it's gonna be a fun night 🙂

@levelsio Mine works just fine. Maybe they made a special EU version that's nerfed. Wouldn't be surprised. I do find it extremely good at pushing down humidity very fast.

Never ever buy a Daikin AC! They limit how many times you can change them to just 150 times per day! We have 7 ACs in the house so that's a limit of 21 instructions per day 😂 completely arbitrary btw, makes no sense, just classic Japanese software design (I didn't buy it, it came with the house, and I've actively been ripping them out and replacing them with Mitsubishi) Other good brands for ACs are LG, Hisense and Haier, can recommend!

Notice how 95% of the posts you see now are about building with AI. Fable this, Codex that. But almost nobody is actually creating a business anymore. Everyone is too busy building and optimizing things that don't matter.

@levelsio Check sunscreen that's reefsafe. Usually it's also natural ingredients, no weird shit.

I've been trying to convince my gf for years sunscreen is cancerous but have as of yet been wholly unsuccessful

@agustinustheoo Yeaaap

@Erwin_AI Wait… So the death of SaaS is mostly because the dopamine rush of hitting more AI tokens and people lacking the other skills to make their businesses viable?

Hot take 🌶 but the "death of SaaS" I think is real. Just not because AI is taking over directly. It's because building is now a commodity. It's easier than ever before. And it feels like progressing your business when AI spits out tons of code. That mindset and perhaps even addiction to building makes it harder for most people to jump into marketing and sales, because it can't be automated as easy and requires much more effort and learning. And so less and less people seem to make bank and become profitable. Which in turn decreases lessons that can be learnt because there's less knowledge how to do this. The ones that know how, are too busy raking in cash. They don't share as much anymore as there's nothing for them to gain/learn either, and they risk being copied by some AI too. All of this reinforces a downward spiral that eventually will cause less people to successfully build SaaS.

@sobedominik I'm taking this personally

i have yet to witness a single beautifully designed SaaS/App that has been designed on a Windows computer prove me wrong