Paolo Micalizzi

Last weekend my mom made an app to run registrations for a half marathon in my hometown in Italy. She and a few local high school volunteers checked in 2000 runners. Signups, bib assignment, race packet pickup, payment, all of it. To build it she used a tool I made called Spunto. It’s a “Claude Code for my mom”. You chat with it and it gives you a link to a web app ready to use or share. My goal with this project is to share the joy I feel building with coding agents with friends and family who haven’t seen this glimpse of the future yet. Seeing my mom have that “aha” moment was exactly it. My cousin used Spunto to plan a trip to Thailand, his roommate made an app to split apartment costs, my wife built a tracker for our 5mo boy’s feeding schedule that we now use every day, and more nice stories like this. This was inspired by how @steipete helped people beyond software engineers “see” what agents are with OpenClaw. Amazing times to be alive. We should try harder to bring more people along this journey. It feels great when that happens. If you are curious: spunto.dev

Claude's first day at Dunder Mifflin

@zhenthebuilder You should show them Spunto spunto.dev

Spending too much time in SF makes you forget how early AI still is Last week in Houston I asked Uber drivers what app they wished existed then had my agent build rough apps for them on my phone before dropoff while people in SF debate how agents will reshape the software industry some of these drivers had never used ChatGPT before a few genuinely thought the apps already existed and I was joking when I said I just made them we still have a long way to go before everyone can fully participate in this shift

@steipete Muahaha

@steipete What’s next on your list?

Seems I have to build all the tooling for the future of software myself. With Claws and Tokens!

You definitely don’t want to be above 800ppm CO2 for a long time if you can avoid it. My agent lets me know when my baby’s room is above the threshold. Suggests to open windows if PM2.5 outside is low. Next step is to automate the air from outside part. Let us know if you find a good solution @levelsio

@mechapreneur @grok Fire is CO, this is CO2 very different

This is Climate Change Hysteria. 8% is deadly, that’s 80,000 ppm CO2 4% is bad 1% is probably annoying, that’s 10,000ppm 1200ppm or 0.12% is normal. We used to sleep in caves with a fire. It’s fine. Hey @grok remind us how much CO2 is in our lungs right before every exhale?

@levelsio @airthings @NorbertDragan You should check this out and build your own 😉 github.com/oloapiu/airshe…

I got the @airthings Plus (unaffiliated, I just like the product), I think @norbertdragan recommended it I been through so many air sensors, and I really think this is the best one so I bought one for living room and then another one for bedroom, and will get another one for my coworking Why so many? Well one of the sensors I bought turned out to be a fake random number generator 😂 Another one kept phoning home to Chinese servers, kinda dodgy. Another one had values that made sense but turned out to be based on kinda estimating from other sensor values, so it didn't actually HAVE the sensor it displayed about (this is common to save money) Why the Airthings is so great: - The device is just super thoughtful and non-invasive, the screen is e-ink (I think?), no backlit, no LEDs shining at you, just black and white, it looks like a paper screen, beautiful, it knows its place! - It measures A LOT of things: AQI (PM2.5+PM10), CO2 (!), VOC, Radon (!), humidity and temperature, and it actually has sensors for all! - You don't need to pair it to WiFi, it just works by itself! (why is this great? So I remember getting that Awair sensor and I was in a hotel nomading and I couldn't even set it up cause captive hotel portal, such an Internet of Shit design to not be able to set up without WiFi) - But when you do pair it with WiFi, it easily connects to your Home Assistant and sends your sensor data to HA without any issue, that lets you automate stuff based on your air quality I love it :D

I agree that people will be busier an more fulfilled than before, as long as we’ll keep getting access to tokens as the models get smarter. My worry is that smarter models will be cost prohibitive for most people and orgs. If a model is smart enough to create outsized economic value, would you not gate it and charge a lot for it? That would accentrate money and power..

i think a lot of people are going to be busier (and hopefully more fulfilled) than ever, and jobs doomerism is likely long-term wrong. though of course there will be disruption/significant transition as we switch to new jobs, the jobs of the future may look v different, etc.

we want to build tools to augment and elevate people, not entities to replace them.

A while ago my wife’s cousin (Taiwanese) WhatsApped me because she was in an argument with her boyfriend about whether gnocchi is a type of pasta. Instead of replying, I made this with Spunto: gnocchi.paolo.spunto.dev I thought it was hilarious

Last weekend my mom made an app to run registrations for a half marathon in my hometown in Italy. She and a few local high school volunteers checked in 2000 runners. Signups, bib assignment, race packet pickup, payment, all of it. To build it she used a tool I made called Spunto. It’s a “Claude Code for my mom”. You chat with it and it gives you a link to a web app ready to use or share. My goal with this project is to share the joy I feel building with coding agents with friends and family who haven’t seen this glimpse of the future yet. Seeing my mom have that “aha” moment was exactly it. My cousin used Spunto to plan a trip to Thailand, his roommate made an app to split apartment costs, my wife built a tracker for our 5mo boy’s feeding schedule that we now use every day, and more nice stories like this. This was inspired by how @steipete helped people beyond software engineers “see” what agents are with OpenClaw. Amazing times to be alive. We should try harder to bring more people along this journey. It feels great when that happens. If you are curious: spunto.dev

@nikunj I don’t think the quality and nuance of Slack’s notification system came from human writing code vs AI. I think it comes from them getting to know their customers well over time…

Every time I see a tweet saying “I can vibe code this in a weekend” - I think of the slack notification system.. It takes time, persistence and effort to get the details right. Sure, a lot of simple workflows will get vibe coded away. And maybe you can put this in Claude Code and get the code right in one shot. But quality, depth and great systems will still have value and take time. You can’t vibe code lessons. Now and forever.

@karpathy Indeed. It was a glimpse into the future

Someone recently suggested to me that the reason OpenClaw moment was so big is because it's the first time a large group of non-technical people (who otherwise only knew AI as synonymous with ChatGPT as a website) experienced the latest agentic models.

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.

@garrytan If you had 1000 Einsteins in a datacenter, would you meter them out or directly set them to generate profit for yourself? Most of us will soon loose access to the SOTA models.

I think it is inevitable that Anthropic and OpenAI eventually roll out $1000/mo and $10,000/mo plans and then reserve the absolute best frontier models to metered access

@melvynx I couldn’t agree more

Day 3 with OpenClaw: In all my tests, GPT 5.4 is consistently the worst model for agentic tasks. Lazy, stupid, never follows anything, feels like you are a baby sitter. I don't know how OpenAI manages to make such a shitty model but this feels terrible. I miss Opus.

@steipete Sonnet is sooo much better than gpt for OpenClaw 😭 It’s day and night.

I'm working on character evals and noticed that Claude would constantly pick itself as #1, so I removed the model names from the judge and changed things.

@pmarca If you had 1000 Einsteins in a datacenter, would you rent them by the hour or set them on the task of making profit for you? Most of us will loose access to fronteer models soon.

Frontier models become more expensive to serve. Token demand explodes at exponential rates. Solve for the equilibrium.

@steipete BYOLLM

The next version of OpenClaw is also an MCP, you can use it instead of Anthropic's message channel MCP to connect to a much wider range of message providers. (I know, this is awkward)

In the future it's not obvious you'll be able to buy AI compute on tap. We're already seeing what's coming: autonomous research loops running ML experiments all night (@karpathy’s autoresearch), zero-human companies that spin up and operate without anyone in the loop, @openclaw style agents managing entire workflows while you sleep. These systems need compute. A lot of it. And the labs building the most capable models will soon face a question: why sell tokens to strangers when we can point 1,000 autonomous Einsteins at the hardest most valuable problems? Even if they keep renting, demand will dwarf supply. Access gets rationed. Small companies get priced out. How else to allocate the profits from autonomous intelligence, other than having flow back to who owns compute and energy? Buying a DGX Spark right now is almost like buying insurance..

@dankuntz I’d be happy to help

@dankuntz @dankuntz interested in adding air quality sensing to it?

Get in loser, we’re making hardware fun again

@levelsio Your perfect competition scenario only applies if your inputs are the same as everyone else's. If your data is genuinely live and proprietary, you're not in that game. You're the infrastructure the $49/month zero human companies would need to query.

If the skill part of making things moves to the AI Then everyone now has access to the same skills So then it's either not about skills anymore and everyone is competing with everyone on equal footing and all of us ending in a perfect competition with close to zero profit So then nobody ends up winning anymor but the AI companies (since we pay them) Or skill is replaced by stuff like ideas, originality, taste, getting users, attention, distribution, audience, capital (who's rich) etc