what a legend.. and this one applies to most things, even tech/AI today
thinking about the human.. in the loop
English
Mark Schellhas
203 posts
Mark Schellhas
@MarkSchellhas
AI Lead at Umain, Eidra
𝕏 Katılım Mayıs 2009
245 Takip Edilen83 Takipçiler
The human brain processes visual information 60,000x faster than text. Humans are visual processors, not text processors. Images hit the brain instantly. Words take work. That's why a single SpaceX launch video communicates more than a thousand-word essay—and why your slide decks hit harder than paragraphs. We're wired for pictures, not prose.
English
@zilasino “Django - The Data Science tool” a little dramatic there 😅 I’ve used Django, it’s nice, but “data science” never crossed my mind. I still use Rails too though
English
Rails vs Laravel vs Django vs NestJS: Why I Still Choose Ruby
norvilis.com/rails-vs-larav…
English
@joemasilotti 🤦♂️ Wake up babe, use Ronel.dev instead, it’s free, and open source.
It’s 2026, this kind of thing should NOT be a paid thing at all (free “until you ship” is not free)
English
Wake up babe, Ruby Native is now FREE for testing!
Joe Masilotti@joemasilotti
Ruby Native is now FREE until you ship to the App Store. Deploy your Rails business as a native iOS app and share it with up to 100 testers via TestFlight. 100% free. - Native tabs, buttons, & OAuth - Push notifications - No Swift, Xcode, or Mac required rubynative [dot] com
English
@yeab2k Good job! I do love how you’re documenting your process. All the very best to you!
English
We killed Google Forms.
Pine AI builds your entire form from a single prompt... questions, branding, themes, analytics. Everything. Instantly.
No dragging fields. No manual setup. No excuses.
It's live today. Try it free. 🌲
pine.roggy.site
English
Unfortunately history shows that it’s not the best architecture that wins, but the popular one (*with momentum). Look at x86, React, PHP, Windows, etc…
We’re humans at the end of the day: we’ll chase memes, follow the herd, pick the widely-accepted tool…
Most devs aren’t engineers, just programmers reaching for whatever everybody else is using.
But forever is a long time, for sure
English
@SimardPete @_chenglou Yeah, it’s hard for me to believe we’ll be using CSS forever
English
This has been a wild morning - I think CSS is dead.
No hype, actually a real engineer, @_chenglou that went through a pretty rigorous process, and in the end, seems to have catalyzed a foundational shift in how websites are made, without CSS.
Decided to publish a special edition of my newsletter that I just released now, because I think this is a pretty big deal. Link in first comment below.
English
English
@karpathy LLM doesn't "argue", it simulates argument. Whatever happened to thinking for yourself? (critical thinking included)
English
- Drafted a blog post
- Used an LLM to meticulously improve the argument over 4 hours.
- Wow, feeling great, it’s so convincing!
- Fun idea let’s ask it to argue the opposite.
- LLM demolishes the entire argument and convinces me that the opposite is in fact true.
- lol
The LLMs may elicit an opinion when asked but are extremely competent in arguing almost any direction. This is actually super useful as a tool for forming your own opinions, just make sure to ask different directions and be careful with the sycophancy.
English
i don't see how @cursor_ai survives - for $200/mo, you can essentially get unlimited usage on claude / codex
that same usage would cost thousands on cursor
even if cursor harness is better, the price difference is just wild
English
@joemasilotti By "production" I mean I have both iOS and Android apps live in app stores - works nicely. Contributors welcome.
English
@joemasilotti Or go cross-platform with Ronel.dev (ronel.dev). I built it for exactly this purpose, for Rails apps. I'm running it in production. And it's open source. My gift to you. Go for it!
English
I had an interesting idea for Ruby Native this morning.
Instead of charging upfront, lets developers use the ENTIRE platform for free. Build iOS apps, submit to TestFlight, upload screenshots, etc.
But toss a little 3 second slash screen “Built with Ruby Native” on launch.
The idea being that serious folks will NOT want their customers to see this. And will then pay. But they get the FULL experience without dropping a dollar.
Lots to figure out first. Will it cost me too much in cloud builds? Add a huge support burden? But I’m hopeful.
What do you think? Would seeing the full picture make it easier to drop a few hundred dollars?
RubyNative [dot] com
English
@ylecun @ClementDelangue @atreides_sf Unfortunately also true beyond AI. Will AMI do it differently?
English
@ClementDelangue @atreides_sf Let's be real, all closed models profit from open models WITHOUT GIVING BACK.
English
well let's be real, open model SOTA was distilled from claude
clem 🤗@ClementDelangue
After @Pinterest @Airbnb @NotionHQ @cursor_ai, today it’s @eoghan @intercom publicly sharing that they’re finding it better, cheaper, faster to use and train open models themselves rather than use APIs for many tasks. And hundreds of other companies are doing the same without sharing. Ultimately, I believe the majority of AI workflows will be in-house based on open-source (vs API). It took much more time than we anticipated but it’s happening now!
English
@karpathy Or could also just use old school frameworks like Rails 😏 it’s CLI generates all the code you need for auth, payments, api, etc - ZERO tokens.
Add kamal for deployment, SSL, etc
Not every problem = LLM
English
When I built menugen ~1 year ago, I observed that the hardest part by far was not the code itself, it was the plethora of services you have to assemble like IKEA furniture to make it real, the DevOps: services, payments, auth, database, security, domain names, etc...
I am really looking forward to a day where I could simply tell my agent: "build menugen" (referencing the post) and it would just work. The whole thing up to the deployed web page. The agent would have to browse a number of services, read the docs, get all the api keys, make everything work, debug it in dev, and deploy to prod. This is the actually hard part, not the code itself. Or rather, the better way to think about it is that the entire DevOps lifecycle has to become code, in addition to the necessary sensors/actuators of the CLIs/APIs with agent-native ergonomics. And there should be no need to visit web pages, click buttons, or anything like that for the human.
It's easy to state, it's now just barely technically possible and expected to work maybe, but it definitely requires from-scratch re-design, work and thought. Very exciting direction!
Patrick Collison@patrickc
When @karpathy built MenuGen (karpathy.bearblog.dev/vibe-coding-me…), he said: "Vibe coding menugen was exhilarating and fun escapade as a local demo, but a bit of a painful slog as a deployed, real app. Building a modern app is a bit like assembling IKEA future. There are all these services, docs, API keys, configurations, dev/prod deployments, team and security features, rate limits, pricing tiers." We've all run into this issue when building with agents: you have to scurry off to establish accounts, clicking things in the browser as though it's the antediluvian days of 2023, in order to unblock its superintelligent progress. So we decided to build Stripe Projects to help agents instantly provision services from the CLI. For example, simply run: $ stripe projects add posthog/analytics And it'll create a PostHog account, get an API key, and (as needed) set up billing. Projects is launching today as a developer preview. You can register for access (we'll make it available to everyone soon) at projects.dev. We're also rolling out support for many new providers over the coming weeks. (Get in touch if you'd like to make your service available.) projects.dev
English
Mark Schellhas retweetledi
After @Pinterest @Airbnb @NotionHQ @cursor_ai, today it’s @eoghan @intercom publicly sharing that they’re finding it better, cheaper, faster to use and train open models themselves rather than use APIs for many tasks.
And hundreds of other companies are doing the same without sharing.
Ultimately, I believe the majority of AI workflows will be in-house based on open-source (vs API). It took much more time than we anticipated but it’s happening now!
English
If only more people knew… I use Grok on a production app, our secret sauce. Better performance (quality + latency), and cheaper
Elon Musk@elonmusk
Grok holds the line
English
Yes, handoffs are changing… next up PRs?
Karri Saarinen@karrisaarinen
Issue trackers were built for an era of handoffs. At @linear, we’re building for an era of agents, context, and execution.
English
@OffGridTech_net @Electroversenet Oh but only if this were true… a lot of “can be” that isn’t being done, and can’t be without more chemicals, “plasma furnaces” (!) and robots…? I’m all for innovation… but green tech doesn’t have a good record in actually doing the “recycling”
English
The oil and gas industry is recycling old propaganda.
Quality solar panels last over 50 years.
Glass and aluminum can be easily recycled. Polymers can also be recycled into harmless molecules by high-temperature incineration in cement kilns with electrostatic precipitators, utilizing embedded energy.
Glass, aluminum, and polymers are the heaviest and most space-consuming parts of solar panels. Once those are recycled, the remaining silver, copper, silicon, and silicone occupy very little space. They can be stored at low cost until automated recycling systems are implemented.
In the short term, as with lead-acid combustion engine starter batteries, a small deposit should be collected to ensure most solar panel components are recycled and the tiny remainder is properly stored until automated recycling methods are implemented.
The hardest parts to recycle may be silicon (semiconductors) and silicone (sealant). Silicon can be recycled with chemicals. Silicone can be recycled into harmless molecules using plasma furnaces.
EV batteries are recycled using a spoke-and-hub strategy: large, heavy components are removed and recycled locally, while remaining materials are shipped to central hubs. Solar panel recycling can follow a similar strategy.
Solar panels are built and installed by robots. They can be removed and recycled by robots. As with EV batteries, functional but less efficient solar panels can be repurposed for an extended useful life to delay the need for recycling. This is already happening.
English
A solar panel's lifespan is just 15 to 20 years. After that, efficiency drops and it needs replacing.
Each panel is built from glass, aluminum, silicone, silver, copper and polymers.
While glass and aluminum can be recycled, the silicone, silver and polymers (the most mining/energy-intensive components) are rarely recovered due to the cost and difficulty involved.
Because of this, the panels simply end up in landfills.
By 2050, solar panel waste is estimated to hit 78 million tons, according to the International Renewable Energy Agency.
This is the dirty secret of 'clean' energy: short lifespan, heavy mining, and a growing toxic waste crisis.
English
It has soon been 10y since the Transformer.
I believe the next dominant AI architecture will be found with AI in the next 1-3 years.
And just like with Attention Is All You Need and the Transformer in '17, initially we won't realize how big of a thing it's.
Change is constant.
English
@OliverMolander Not exactly tech “entrepreneur” more than tech “innovator” … don’t forget Alan Turing and Tim Berners-Lee… where would Linus be without them?
English
Name one Nordic (or European for that matter) tech entrepreneur who has been more influential globally than this guy.
I'll wait.
🇫🇮🐧
English