2026-04-01

Latino Task Force and Calle 24 are both bully orgs that think they “own” the Mission. They bully non-Latino businesses. They’re desperate to keep Jackie Fielder in office so they can keep the grift going.
MissionLoco: SF’s “nonprofits” are rife with corruption. In the Mission, a web of overlapping nonprofits (many are “zombies” dusted off as $$ conduits) pass the money around. Loco Bloco, Homie, Precita Eyes, Latino Task Force, Mission Education Project, Calle 24, and on and on.@DanielLurie

Re @garrytan AI is giving the olds a second wind
Hayley: The majority of @atlas startup founders are over 35 years old for the first time. We’ve seen a surge in startup formation across all age groups, but the over 35 set has grown the fastest.

I do Sunshine Act so you don’t have to!
It should be a much bigger deal that City Attorney is investigating a leak of a confidential document *from the Board of Supervisors*!
I got this letter to BoS from Chiu, where he explains why *removal from office* is on the table.
More:

The research behind this is wild. If you played Pokémon as a kid, you have a tiny region in your brain that exists only because of Pokémon. Not a metaphor. Stanford put people in brain scanners and found it.
The study was published in Nature Human Behavior in 2019. They scanned 11 adults who grew up glued to their Game Boys and 11 who never played. When they showed both groups images of the original 151, the players' brains lit up in one specific spot every time. Same spot across all 11 people. The non-players showed zero response.
That spot is a little fold in the back of your brain that normally processes things like animal shapes and cartoon faces. In the Pokémon players, a chunk of it had been permanently reassigned. Their brains carved out a Pokémon department sometime around age 6 or 7 and just never took it down.
And the reason it ended up in the same place in everyone's brain comes down to the Game Boy itself. The screen was 2.6 inches. Every kid held it at roughly the same distance. So those 151 characters hit the exact same patch of each kid's retina, thousands of times, during the years when the brain is still soft enough to reorganize itself. Where an image hits your retina in childhood is what tells your brain where to build the wiring.
Reading works the same way. Humans invented writing about 5,000 years ago. There's zero evolutionary reason for a brain region dedicated to recognizing words. But every person who learns to read grows one, roughly the size of a dime, in the same part of the brain.
Brain-imaging research from 2018 actually watched it appear in children's heads as they learned their letters. It grew by quietly taking over nearby tissue that wasn't doing much yet. Stanford published a follow-up this year showing this region is way smaller or missing entirely in kids with dyslexia, and that 8 weeks of intense reading practice physically grew it back.
London taxi drivers show the same thing in a completely different part of the brain. Brain scans from a 2000 study found the region that stores mental maps had physically expanded, and the longer they'd been driving, the bigger it got. These drivers spend 3 to 4 years memorizing 25,000 streets before they get licensed. About half wash out.
The common thread is childhood. Harvard researchers trained young monkeys to recognize new shapes and they developed brand-new brain regions in predictable locations. Adult monkeys trained on the same shapes never got those structural changes. The young brain wires itself in a way the adult brain cannot replicate.
If you're wondering whether a Pokémon patch in your brain means you lost something else, no. The region sits alongside your normal visual processing areas, not on top of them. Your brain has hundreds of millions of neurons in that zone alone. The lead author noted that every participant in the study had gone on to earn a PhD.
Fanatics Collect: A Stanford study found that people who played Pokémon heavily as kids developed a small region of the brain that responds specifically to Pokémon characters.
Researchers scanned adults who grew up playing on Game Boy and showed them images of Pokémon like Pikachu and Bulbasaur.

Has anyone used Matic (robot vacuum)? If so, from 0-10 (no 7 allowed), how strongly would you recommend and why? Thanks! 🙏

Claude Code 源码泄露事件后续越来越精彩了。
有人拿泄露的源码丢给 OpenAI 的 Codex 分析，竟然找到了 Claude Code 疯狂消耗 token 的元凶——autoCompact（自动上下文压缩）机制在失败后会无限重试，完全没有上限。据源码注释记录，曾有会话连续失败高达 3272
次。
修复方法简单到离谱：加一个 MAX_CONSECUTIVE_AUTOCOMPACT_FAILURES = 3 的限制，连续失败 3 次就停止重试。三行代码，搞定。
打完补丁后，这位老哥表示使用额度恢复正常了——之前被吐槽的"用两下就触发限速"，很可能有一部分就是这个 bug 在背后偷偷烧 token。
仓库地址放下面了。
Lydia Hallie ✨: We're aware people are hitting usage limits in Claude Code way faster than expected. Actively investigating, will share more when we have an update!

been using @garrytan's gstack office-hours for quite a bit now.
it now gives me relevant yc youtube vids and PG's essays based on my situation.
it also ruthlessly cuts off any bs if i'm about to spiral into my thoughts.
pretty g now

That’s why I usually say that comparing open models with closed-source APIs or products is like comparing apples and oranges. Or comparing an engine with a full car. Or comparing an ingredient with a Michelin dinner (missing ingredients, prep and chef).
There’s a lot of scaffolding and tricks that is done behind APIs that make it unfair to compare to a raw open-source model! However that also means that if you put the work to really try to make open-source models work, they can give you much better results than what benchmark suggests (and that’s without even post-training/fine-tuning)
Sebastian Raschka: http://x.com/i/article/2038978163389112321

Brian Conrad found rampant citation misrepresentation in the California Math Framework. Now we see there was also an effort to cherry-pick results to report. 😞
Rahim Nathwani: Jo Boaler emailed her 'detracking' mailing list asking for examples of schools that detracked. But only if the results were 'positive.'
That's not research. It's shopping for evidence.

http://x.com/i/article/2039243168689139712

The people who ship 100 things aren't braver than you. They just got bored of their own self-importance faster.
Kpaxs: A child learns to walk by falling. Founders learn to build by launching things that don’t quite work. Writers learn to write by producing drafts they later hate.
If you fail a lot in public and keep going, you can’t cling too tightly to the idea that you’re special.
You’re

I agreed to get a Phil Foden haircut if 17 yo gets 1550 on the SAT. For people outside the UK, this is the risk I'm taking here:

The most AI proof job in the world is entrepreneurship
Use it to make products and services. Build more companies. On Shopify or otherwise.

Here's my new tutorial on how to build a mobile app from scratch in 16 minutes with Claude Code, @pencildev, and @expo:
1. Define requirements with Claude
2. Create beautiful designs with Pencil
3. Build the app with Claude Code and Expo
Watch it to build a mobile fitness app complete with live workouts, progress charts, and more.
📌 Watch now: https://youtu.be/oS53by4Hwvo

This project is insane and most devs still don't know about it.
A full Chromium browser that runs entirely in your terminal.
→ supports WebGL, WebGPU, audio, video
→ starts in less than a second
→ runs at 60 FPS
→ idles at 0% CPU
→ works through SSH
→ no window server needed
→ you can literally watch YouTube in your terminal
it's called Carbonyl, 15.9K stars on GitHub
built with Rust btw.
100% open source.
http://github.com/fathyb/carbonyl

awen (@awen_ai) translates intent into exceptional visuals through conversation.
Congrats to the team on the launch!
https://www.ycombinator.com/launches/Ppr-awen

Introducing the "Follow builders" skill: the best way to stay on top of the insane happenings in AI
I carefully curated 25 X accounts & podcasts that share the highest-quality, first-hand insights on AI (by builders from OpenAI, Anthropic, Google, OpenClaw, Replit, Vercel, Cursor...)
Your OpenClaw/agent can remix my central feed & send you a personalized daily newsletter in whatever channel you like
Already widely used with 2k+ stars on GitHub; link below

California housing policy for the homeless (Housing First) "recognizes drug use as part of people's lives" the result? 30% of all OD deaths happen inside supportive housing. SB1380. One of the worst laws passed in California history. @CAgovernor @auweia1

this is insane AI design alpha from google:
google's lead stitch designer just showed how he turns a 1-line prompt into a site that looks like an actual design agency built it.
he walked through every decision, from the first vague prompt to a finished design with real content, real layout, real direction.
here's his full process:
the problem is most people open stitch and type something like "a road running race listing page."
stitch gives them a generic dark layout. it works, but it doesn't feel like anything.
it's effectively AI slop
his approach starts somewhere most people skip entirely.
1. he starts with empathy.
before touching any tool, he asks:
> who is this site for?
> how should they feel when they land on it?
say he's building a community marathon site.
these are historic races in world-class cities.
the site should feel prestigious, like standing in the jefferson memorial.
so now he knows the feeling he's going for.
but stitch can't build from "prestigious." it needs design language.
2. he asks gemini to translate that feeling into words stitch can actually use.
instead of vague words like "sporty" or "athletic," gemini comes back with phrases like "architectural limestone," "ink on paper," "clay on an old track."
he feeds those into stitch and the output jumps immediately.
real structure, real intent, something you could actually work with.
so now the direction is set. but the default colors and fonts stitch picked don't match the feeling yet.
3. he dials in the design system (the set of colors, fonts, and components that keep every screen consistent).
he doesn't think of colors as a matching palette. he thinks of them as a hierarchy. each one has a job:
> neutral is the canvas. ~80-90% of the screen. he sets it to warm architectural limestone
> primary is the ink. headings, body text. he drops it to dark asphalt
> secondary is more subdued so primary text keeps focus
> tertiary is the accent. loudest color, used least. he sets it to a clay red that pulls your eye straight to the call to action
for fonts he picks public sans. official but friendly.
like a prestigious journal you'd actually want to read.
so now the colors and type feel right. but the layout is still generic.
4. he fixes layout by thinking about physical objects.
"if my website was a book, what kind of book would it be?"
his answer: a coffee table book. full-page imagery, dense info, editorial headings.
he uses variants (a stitch feature that generates multiple layout directions at once) to explore editorial lookbook layouts with large typographic headings.
like a luxury travel magazine.
so now the layout, colors, and typography all feel dialed in. but scrolling through, something still feels off.
5. the content.
the headings say things like "the elite calendar." the aesthetic is there but the words are generic. it doesn't feel like a real site yet.
so he installs a copywriting skill (an agent instruction file with expertise in writing web copy)
and feeds it all his context plus the design .md (the creative brief stitch auto-generated from his prompts).
the skill drafts page copy, he reviews and edits, then pastes the final version back into stitch.
now the site has real names, real tips, real ctas. it stopped feeling like a template.
the whole process: empathy → design language → colors and type → layout → copy.
if you've been getting generic output from stitch (and AI design tools in general), start with one question:
how do you want the user to feel?
everything else follows from that.
Stitch by Google: We are completely humbled by the amazing response to our launch last week! 🫶 Now, we want to help you get the absolute best results from Stitch.
In this new video, David East walks you through how to consistently get premium results.
We also launched a new prompt enhancer

Social housing will never be good and no one will want to fund it until @sfgov bans drugs and trafficking. Otherwise they’re just tax funded crack dens. Social housing should prioritize young families.
T Wolf 🌁: California housing policy for the homeless (Housing First) "recognizes drug use as part of people's lives" the result? 30% of all OD deaths happen inside supportive housing. SB1380. One of the worst laws passed in California history. @CAgovernor @auweia1

It’s a very good point that S.F. politics gets personal and intense.
For example, one time the Supervisor in this article posted my face and full name on her Instagram and called me fascist.
Because I criticized her zoning proposal for laboratory use.
The San Francisco Standard: Local officials speak out about pressure, stigma, and survival after Supervisor Jackie Fielder's hospitalization, which led her to publicly consider resignation.
📝: @FitzTheReporter https://sfstandard.com/2026/04/01/sf-politics-fielder-mental-health/?utm_campaign=trueanthem&utm_medium=social&utm_source=twitter

It took @nityeshaga & I well over 100 hours to build our first agent, Claudie.
Then, it took 7 hours to build Jean-Claude, our sales pipeline manager.
⬇️ Nityesh shares what we've learned in the process of building our agents and the architecture that has worked best for us. Our next teammate in training is, of course, an @every Plus One.
https://every.to/p/what-i-learned-onboarding-our-ai-project-manager

Today
We're launching Skill-Chaining
Slate is able to choose which skills to use and the order in which it uses them
Chaining them together into more complex behaviors
This is an experimental release, but hope you enjoy!

http://x.com/i/article/2038869305555824640

We experimented with AI internally, but took a deliberate approach in the product, understanding which workflows mattered and waited for the models to improve.
That meant giving up being first, but it gave us a much clearer view of where AI can actually be useful for companies
Dan Shipper 📧: SaaS isn’t dead, it just needs to become agent-native. Linear (@linear) is a great example of how:
They pivoted the product to be used by both humans and agents, and that has made them one of the premier software tools in the agent-native era.
I had Linear’s cofounder and CEO

Y Combinator CEO Garry Tan’s advice for startups: “When you’re small, act small”
A lot of founders try to emulate large companies, and will do things like use the same terminology as Microsoft to describe their products. But Garry argues this is a mistake:
“When you’re starting something new, the whole advantage is that you’re a real human being. We are so starved for real, authentic connection that if you can talk to people and say ‘Hey, I’m the CEO. What do you need?’ That’s the most powerful thing.”
Being small lets you offer fanatical customer support. Not only will this win customer trust, but it’ll help you find product/market fit. If you listen to customers, they will tell you what they want.
“The reason why people don’t do this is they think starting a startup is building this incredibly complex machinery… But I encourage you to think about it in a different way. It’s more like throwing a really, really amazing party… You go there, you see a friend, they say ‘Welcome! Let me take your coat. Let me introduce you to your friends.’”
For his first startup, Posterous, Garry and his team aimed to reply to every single customer support email within ten minutes. And if there was a bug, they fixed it on the spot.
Human connection with your customers is really important. Garry cites a study on Usenet that found retention increased from 16% to 26% if someone received a reply to their post on the forum.
As Garry explains:
“A 10% difference in retention is actually the difference between a startup that’s flatlining and one that’s working. The compounding of this is really, really massive… Be small. Be human.”
Video source: @ECorner (2023)

Tomorrow on Long Strange Trip: I get to talk to @jack on how to run companies in the AI era, joined by @roelofbotha

Slate is one of the more ambitious projects out there right now. Find myself consistently agreeing with the direction they are taking. Really smart team. Bullish!
akira: http://x.com/i/article/2038869305555824640

We’ve had a monster first four weeks.
In a major narrative violation, 56% of our donations were under $300.
This is a grassroots movement.
The energy to replace the hypocrisy and selfishness of @RoKhanna is palpable.
I’m immensely grateful to every donor and to the tireless Agarwal for Congress team.
Still a long ways to go to! Let’s go. 💪🏽💪🏽💪🏽

Having done YC before, it's amazing to be back. The energy is the same 14 years (!!) on. Also, the quality of the folks I met was just as a high, if not more. So many have done a ton already, some are early. All are moving super fast.
Crazy achievement by the YC team and @garrytan.
My personal takeaways from the kick off of P26:
- the vibe/how folks/users feel matters - prolly even more for hardware
- what is the smallest thing one can actually ship rn
- watch out for hero-mode; are we solving the real problem, or the easy one
- strong cofounder bond is key, and the easiest way to fail, even vs funding
- building a pro sports team, not family
- post more to LinkedIn, as spaced rep is a thing, and starts a brand
- create a tribe. convert people. if founders don't no one will: but build one tribe, not more

Coolest piece of investment memorabilia I’ve ever seen
Sequoia Capital: In honor of 50 years of Apple, we're sharing - for the first time ever - Don Valentine's original 1977 memo for Sequoia's investment into Apple Computer. #Apple50

> “The Marina is for Everybody”
> *protesting new housing in the Marina
ConnectedSF: Happening Now: ➡️Protest opposing 25-story Marina Safeway Development proposal. Last week, SF Planning Dept deemed eligibility of AB 2011 approval, fast-rtacking design review.
Fight back by signing the petition that will be emailed to City Hall today
https://form.jotform.com/253625476102151

.@LiamFedus brainmogging and framemogging at the same time is crazy work
atlas: the legends at @soontechnology and I have been on a mission to find the most jacked AI researcher and we think that @LiamFedus, co-creator of ChatGPT and co-founder of @periodiclabs might be it.
check out the latest episode of Swole as a Service where he totally framemogs me

NEW: SF Supervisor Jackie Fielder checked herself into a mental hospital last week after possibly leaking a confidential city memo (see: very illegal). Is our favorite Waymo-hating socialist politician hiding something? @EvanMilenko investigates 👇
https://www.piratewires.com/p/did-san-franciscos-marxist-supervisor

"We treat software like it is so precious. And in the new world we're going to, it's going to be much more fungible. But that's the good news."
@garrytan on where the moats will be for future YC companies:
"That's where taste, agency, and trust matter a lot. This is something that's only crystallized recently. We've been talking about agency and taste for the longest time. Agency is being able to prompt, and taste is being able to do evals."
"Trust is the third thing. It's probably the most important. When you have an enterprise company that is selling to real businesses that are relying on you — that's your moat. Your moat is actually trust. They went through the effort to try you, get you onboarded, and incorporate you into how they work."

California Teachers Association kicks in $5 million to the committee backing an initiative to make permanent the 2012 Prop 30 tax increases that were sold to voters as 'Temporary'.
CATargetBot0001: NEW F497
California Teachers Association/Issues PAC
$5,000,000 To CALIFORNIANS FOR PROTECTING PUBLIC E...
http://cal-access.sos.ca.gov/PDFGen/pdfgen.prg?filingid=3130356

Look mama, @CoraComputer has an iOS app with Push. Also the great @joemasilotti is helping me because I'm not as good as he is! Love him and his RubyNative
DM me if you want alpha access

20 million developers used to be the gatekeepers to software.
Wabi CEO Eugenia Kuyda on who gets to build now:
 "The only people who could make software up until last year or so [were] just professional developers... Very few people in the world."
"Even if you have a good idea, go ahead, find an engineer, find a co-founder, find a technical co-founder to build this. But you can't really otherwise build it... It requires a lot."
"But now anyone can."
"This is just wild."
"Before  we had to spend months developing that, developing an app and figuring out some illustrations and sounds. Now you just write one prompt and it's there."
@wabi CEO @ekuyda on @solofounders with @julianweisser