Suhail
The Chinese do not need to distill anymore. They've climbed up the proverbially tech tree that to climb higher, they need to do actual frontier work. Distillation is marginal.
Any lobbying or legislation that calls for banning open weight models in the name of "distillation" is total BS. This is a fight against future American innovation.
The labs know that open weight models completely commoditize their revenue generator. They know that we will optimize it to 1/100th the costs. They know that we will stop concentrating our workloads on them. If we start using open models, they lose the understanding of workloads that they can build products around to *replace* your products as they move up to the app layer.
Remember: *they* trained on humanity's data and didn't pay a cent. If you own intelligence, you can tax humanity.
Do not let them fool you.
My 8-year-old and I built a @ChatGPTapp Site to help her learn the multiplication tables.
She loves animals, so we:
→ Used ChatGPT Images to generate the UI and characters
→ Added some awesome music
→ Even added a timed boss level
📌 Play it here: https://serenas-safari-math.petergyang.chatgpt.site/
ChatGPT: Create something worth sharing with ChatGPT sites.
Share your site with us for a chance to win a limited-edition swag box, including the sold out ChatGPT basketball 🏀
Giveaway closes this Sunday at 11:59PT, winner will be chosen randomly.
Happy building.
Suhail
Re They *distilled* humanity. The hypocrisy is unreal. The best gift is to give it away, at the lowest marginal costs, to let us prosper as a species. If intelligence casts a light cone, it is a moral imperative to extend it as diffusely as possible.
The Standard probably should try to stick with more journalism and less partisanship
Disappointing how journalism is sometimes weaponized like this
It’s not unreasonable to ask for accountability from our cities and attacking accountability is speaking power to truth
Debora Allen: Congrats @marcjoffe, @cocotax, SHIFT-Bay Area, and http://transitaccountability.com — you’ve officially earned the “SF Media Hates You” badge 🏆
@sfstandard just dropped an early hit piece that’s so sloppy it’s almost impressive. They spent years rightly calling out @SFBART’s waste,
Francois Chaubard
agreed.. magnus carlson level psyops to get ppl to believe this is how you cool a gpu and write about it..
djcows: i have nothing but respect for whoever created the AI datacenter water psyop
it must have required an enormous amount of planning to trick that many, often smart, people into believing water is an actual issue - absolute peak social engineering chess
Based on internal evals:
▪️ Kimi K3 is top-tier at cybersecurity
There is chatter on X that Moonshot benchmark-overfit. These are stealth evals. Model has raw IQ.
▪️ Sol is a leap ahead in cyber capability
At a significantly higher cost, but quite remarkable still.
▪️ Fable refuses everything
We couldn’t get it to complete the run at all. What’s interesting is that Sol in comparison was much more open to helping with defensive cyber hardening
TL;DR: frontier, open-weight cybersecurity capability is here. Try it on http://deepsec.sh for defensive purposes.
Malte Ubl: We ran Kimi K3 on a private cybersecurity benchmark.
TL;DR: Kimi K3 is the workhorse for cyber security tasks at great recall/precision/price. GPT 5.6 is best recall/precision but at 7x higher cost per run.
For context, http://Deepsec.sh is an open-source cyber harness
Morgan J. Freeman
🚨🚨🚨 BREAKING: In a fully corrupt & desperate move, Trump is forcing the Postmaster General to REFUSE delivery of mail-in ballots for the 2026 midterm elections to any state that does not comply with his so-called “election integrity” executive order.
Let that sink in… and let the lawsuits begin.
Source: @OccupyDemocrats
Cris
😳😂🤣🤣🤣
evan conrad
anyone want to help fund getting this exact billboard up on the highway in sf
Kane 謝凱堯
To compare the systems, consider the fate of anticapitalists under capitalism (consulting revenue) versus anticommunists under communism (slave camp).
levy: what does capitalism mean
Actually fund recovery and treatment
Heavily audit and defund any harm reduction organization that actively encourages, aids, and causes overdose deaths and supports drug dealers
It’s very easy, SF just needs the will to do it
T Wolf 🌁: This is how crazy the harm reduction people are. This is who we're up against. It's people like this who must be de-platformed and removed from drug policy. More importantly, it's people like this who were calling the shots prior to Drug Czar @SaraCarterDC. @ONDCP
Socialists want state controlled bread lines
You know what creates abundance? Great functioning markets.
Matthew Yglesias: You could have a society where everyone lives in public housing but you would still “have a landlord,” it would just be some kind of public agency.
Jesus
Brutal
Yann LeCun
Re @hosseeb @deanwball Soooo, releasing Linux was dumping?
Apache, MySQL, PHP?
HTTP, TCP/IP, OpenSSL, OpenSSH?
Libjpeg, VLC?
The open source software stack of the mobile communication network?
Signal?
PyTorch?
Llama?
GSkills anyone?
brett goldstein: my hot take is that 99% of software can be built with the same set of finite elements
spending time/tokens building any of this from scratch over and over again is peak AI psychosis
Markdown files are universal and will survive millennia
Unsurprisingly they are a fantastic data format when the intelligence stack is in flux at this extreme pace
kepano: formats evolve over decades, apps over months, intelligence is evolving every day
in terms of pace layers "file over app" defines priorities: invest deeply in formats, and don't expect anything at the intelligence layer to last very long
Daniel Jeffries
Re Dumping? Come the fuck on. Stop trying to use clever language to make your points and make them honestly @hosseeb
This is a just blatantly transparent dog whistle to the administration which used similar language over solar panels and cars with China.
Was Android dumping?
Open Source has a long history of commoditizing a layer to play to the strengths of the culture/capabilities/skills/business of a rival. It's a loss leader. You give away a code piece and build products and services around it. Google used their strength in scale and manufacturing and partnerships to out maneuver Apple.
Apple is still doing just fine, last I checked, and making a premium product.
Red Hat, MongDB, Hashicorp and other billion dollars open source companies did the same.
As Yann wrote, was Linux dumping? Postgress? Docker? MongoDB. Etc.
Driving down the cost of goods and services in an economy is good so that we can use more of that service in more innovative and interesting ways.
You don't get the modern internet built in IIS server. You get it on Apache and NGINX.
Last I checked Linux powers more than 50% of the Azure cloud now. Imagine if Microsoft had been successful killing open source with their short sighted early idiocy when they called Linux "communism" and "cancer" in another ham fisted attempt at rhetorical flourish? They would have shot their own future core business in the face!
We are such a county of short sighted people with complete lack of long horizon thinking now.
The dumping frame is straight nonsense and what's worse is I bet you know it and you just do care and decided to use this sleight of hand language anyway as a dog whistle to short sighted fools.
Bet on every model at once
http://vercel.com/ai-gateway
Peter Yang
“We’ve got /loop, /goal, and workflows. They all get the agent to run for long periods of time.”
Here’s my new episode with @trq212 from the Claude Code team, where he shared how he plans, builds, and runs loops with Claude Code:
→ How /loop, /goal, and workflows work differently
→ Live demo: Editing launch videos with Claude
→ Using HTML artifacts to plan and learn
Some quotes from Thariq:
“We cut Claude Code’s system prompt by 80%. As models have gotten smarter, they need less direction, fewer constraints, and fewer examples.”
“Planning is an iterative process of exploring, investigating, and finding out what you don’t know.”
“One failure mode I see is that people glaze over AI’s plans. You want to make sure it’s something you read.”
I've been wanting to interview Thariq for a long time and I learned so much from this epsiode.
📌 Watch now: https://youtu.be/aVO6E181cNU
Thanks to our sponsors:
@RiversidedotFM: All-in-one AI studio for podcasts and video https://creators.riverside.com/PeterYang
@WisprFlow: 4x faster than typing with your voice https://ref.wisprflow.ai/peteryang
POV: you’re the Messi of motorsport 🏁
Anthropic recently cut Claude Code’s system prompt by 80%.
@trq212 explains why:
“As the models have gotten smarter, they need less direction, fewer constraints, and fewer examples.
The examples are constraining it because now it’s like, ‘Oh, you want things like this example.’ If you remove the examples, it can actually be more free-form.
All this is to say that you want to trim your context [when a new model is released]. The latest models often need more room to run.”
📌 Watch the full episode here: https://youtu.be/aVO6E181cNU
Peter Yang: “We’ve got /loop, /goal, and workflows. They all get the agent to run for long periods of time.”
Here’s my new episode with @trq212 from the Claude Code team, where he shared how he plans, builds, and runs loops with Claude Code:
→ How /loop, /goal, and workflows work
Peter Yang
Anthropic recently cut Claude Code’s system prompt by 80%.
@trq212 explains why:
“As the models have gotten smarter, they need less direction, fewer constraints, and fewer examples.
The examples are constraining it because now it’s like, ‘Oh, you want things like this example.’ If you remove the examples, it can actually be more free-form.
All this is to say that you want to trim your context [when a new model is released]. The latest models often need more room to run.”
📌 Watch the full episode here: https://youtu.be/aVO6E181cNU
Peter Yang: “We’ve got /loop, /goal, and workflows. They all get the agent to run for long periods of time.”
Here’s my new episode with @trq212 from the Claude Code team, where he shared how he plans, builds, and runs loops with Claude Code:
→ How /loop, /goal, and workflows work
gm 🇦🇷
Thariq
working more on a post about what we learned doing this and how you can apply this to your skills and system prompts
Peter Yang: Anthropic recently cut Claude Code’s system prompt by 80%.
@trq212 explains why:
“As the models have gotten smarter, they need less direction, fewer constraints, and fewer examples.
The examples are constraining it because now it’s like, ‘Oh, you want things like this
evan conrad
this post is so polarizing because people literally seem to think its impossible to make prices go down; this city literally builds superintelligence but cannot imagine upzoning to minor european levels of density
evan conrad: anyone want to help fund getting this exact billboard up on the highway in sf
I'll be honest this cat scares the crap out of me for this match
one of the best features of ChatGPT Work is that it runs in the cloud, meaning that it works from mobile, with your laptop closed.
kinda crazy how long the main way to get the magic of agents has been while leaving your laptop cracked open!
Marcel Haas
With @Replit u can easily scale your app workloads across multiple dedicated machines with different roles (Worker, Webserver, Static).
Let's say u want to offload a long running and compute intense batch operation to a dedicated resource, u can just use a worker type resource for that.
The agent helps to distribute, orchestrate and configure everything. Replit is not just code generation, it's a advanced platform for your business apps from secure infra to code.
How can a 19 year old kid be so good
This is going very poorly for Argentina
Justin Liao
this is by far the most beautifully messy personal websites I've seen so far.
i'm inspired @eve_bouff
gonna update mine soon!
Christopher Webb
Imagine declassifying documents to prove your claims, only for them to do the exact opposite. Everything Trump does is half-assed.
You just know this dude cheated his way through school.
DOCUMENTS REVEAL
📌 Russia identified as the primary country attempting to target election-related systems
📌 The “stolen” election data was publicly available and legally available for purchase
📌 Intelligence experts concluded that widespread manipulation of election outcomes unlikely
📌 The greatest threat identified is the spread of false claims designed to erode public trust in election integrity
This is the best analysis I’ve seen so far.
Martinez is MVP in this match for sure no matter who wins
Argentinian goalkeeper sure is earning his salary.
This is pure torture
Let’s go España!
Thankful to 🇦🇷 Argentina for once again giving it all. Grit, passion, talent, faith. To Messi for proving he’s the best ever. And to America 🇺🇸 for hosting such a wonderful event. Enhorabuena tíos 🇪🇸!
Feels like a watershed moment for advancing mathematics. Scientific and medical advances which can really improve people’s lives feeling very close now.
Thomas Bloom: So far GPT 5.6 Sol is the most interesting new model mathematics-wise for me - those of the new proof claims on http://www.erdosproblems.com that I've looked at in detail have all been correct, and moreover contained some interesting ideas.
1/7
David Sacks
Kimi K3 just fixed 15 critical security bugs that Codex and Fable refused because of “cyber guardrails.” There’s no reason to limit American models on tasks that Chinese models handle without issue. We’re only making ourselves less competitive.
calle: I have a report full of security issues of a software I'm working on.
Codex won't fix them because of Cyber guardrails
Fable won't fix them because of Cyber guardrails
Kimi K3 fixed them all. No restrictions, just gets the job done.
This will end badly for OpenAI & Anthropic.
