Samantha Power
Starting in early 2025, Elon Musk and the Trump administration began terminating USAID's programs and firing its staff — with Musk himself boasting about "feeding it into the woodchipper." One year ago today, USAID was officially dissolved, its remaining programs haphazardly folded into the State Department. Amid all the lies and misinformation that have followed, some facts about what has actually been lost:
• USAID saved more than 3 million lives a year at a cost of less than $10/month per American. That is what was destroyed. On purpose.
• According to Boston University's Global Impact Counter — which tracked deaths attributable to the cuts until it stopped operations in February 2026 — an estimated 781,000 people died preventable deaths in the first year, including 518,000 children.
• Global child mortality (the number of children who die before their fifth birthday) rose in 2025 for the first time in 35+ years — by 200,000 additional deaths.
• USAID's 50-country disease surveillance network — the system that cut outbreak response times from 2 weeks to 48 hours — is gone. We are now watching an unprecedented Ebola outbreak unfold in real time — with the highest first-month caseload and death rate in modern history.
• Programs reaching 93 million women and children were cut 92%. TB programs cut 56%. Water and sanitation cut 86%. Over 2,000 health facilities permanently closed.
• 25 million fewer people received humanitarian assistance in 2025. The overall humanitarian budget was slashed 74% — from $14.1 billion to $3.7 billion.
• 363 million people face acute hunger in 2026. The famine early-warning system that would have seen it coming went dark for five months.
• $1.7 billion in democracy and governance funding (election monitoring, anti-corruption work, support for independent media and civil society) was terminated.
• 360+ independent media outlets lost funding. Hundreds of legal clinics closed.
• Far from saving money, the Trump administration itself has already said the dismantlement will cost taxpayers at least $19.2 billion in cancellation fees, severance, and penalties. That's more than half of USAID's annual budget — spent on destruction and closeout, not support for vulnerable people.
• American farmers, universities, and businesses are among the casualties too. USAID partnered with more than 3,500 U.S. companies and maintained 17 university-based research labs. Its work with U.S.-based contractors and the private sector generated hundreds of thousands of American jobs and multiplied the return on every dollar spent. Those markets and partnerships are gone.
Mike Schroepfer
Did a long chat with @yayitsrob on Shift Key about the Gigascale thesis, hard tech, and why the "US can't build anything" take is wrong.
A few things I believe: 🧵
RT @davidaxelrod: In his SOTU speech in 2010, @barackobama warned that with it's Citizens United decision, the SCOTUS had "opened the flood…
Ashley Ha
Don’t sleep on AIE guys
Ashley Ha: Don’t sleep on Codex guys
@aiDotEngineer
Matthew Berman
I’ll be at AIE tomorrow. I’m doing a panel on local AI and then a live podcast with @swyx.
Come say hi!
Theo - t3.gg
Kinda want to do a meet & greet at AIE tomorrow. Who wants to host me at their booth? 👀
Democratic Wins Media
BREAKING: CNN just aired a devastating montage of all the times that Donald Trump accused others of using their public office to make money. Donald Trump's net worth has almost tripled since he took office. Everything he accuses others of is an open admission.
As someone still trying to learn how to read code this is great! Installing the explain-diff skill asap
Geoffrey Litt: Hot take: I think it's still important to understand the code that our agents write!
In this mega thread (based on my AIE talk today), I will explain why that's the case, and show some ideas for how to efficiently understand code. Alright, let's dive in. 1/
Shawn Presser
I’ll be actually-homeless soon (living out of my car) if I can’t land a job. Our house is being sold within two months if I don’t find work. I’ve been trying everywhere, but very few companies are answering. So please, email me or DM with literally anything involving computers, or ask a friend, or your business’s HR person. (I’d love to work with you.) I’ll do a good job, and I can learn and adapt to any working style you need.
I was employee #2 at Carmack’s AI lab, Keen. I worked at Groq helping to design and implement their LPU chips. I made Books3, a dataset trained on by LLaMA and Claude. I’ve been programming for 25 years.
Location: Lake Saint Louis MO
Remote: Yes
Willing to relocate: yes
Technologies: Python, PyTorch, JAX, C++, JS/TypeScript, Objective C, Swift, Rails, Linux, Lisp, and anything else you give me a week to learn.
Résumé/CV: https://shawwn.net/docs/shawn-presser-resume.pdf
Email: shawnpresser@gmail.com
I wanted to avoid the humiliation of begging, but nothing is working. I need money or else our 3yo daughter loses her birth home. I’ve been through dozens of interviews from the companies that do answer, but the market is bad right now. You can have me for a steal.
Sarah Chieng
Token billionaires lounge
must apply and submit token spending reciepts
@_lopopolo @swyx @communidiyi @jxnlco
Jenna Pederson
The question is no longer “what can the agent do” because the list of things it can do keeps growing. The real question we should be asking is: what can only a human be accountable for? @addyosmani at @aiDotEngineer
T Wolf 🌁
These politicians will pull their endorsement over "allegations" while Manny's opponent, Gary McCoy has multiple felony convictions. Let's see if they endorse him. San Francisco politics is seriously special ed.
Adish Jain ☕️
thank you for the book @garrytan @ycombinator
heard great things about this book by @markpinc
for what it's worth, i only invite double-length track keynotes when I'm very sure that both speaker and content deserve it. Today, @chrmanning and @abshkbh did double duty at AIE and by all accounts* people loved the opportunity to go deeper on sandboxing and world models. Look at this insane room - and the online audience is going to be >1000x this!!
*i unfortunately have to do show duties so rely on secondhand accounts
swyx @aiDotEngineer WF: i havent watched all the online talks yet but am binging this one now and it is exceptional.
we are very lucky to have all this sandboxing teaching for free. meet abhishek at aie today! he’s roaming around!
Tyler Black, MD
Since Trump took office:
* Elon net worth is $980B (+$500B)
* Donald Trump's family ~$10B, up about $7B
* Top 15 billionaires: up 33%
And USAID was cut 6 billion, resulting in hundreds of thousands of or the poorest people needlessly dying and suffering.
/1
Latent.Space
Today at the AI Engineer World's Fair in San Francisco: the 'software factory' vision met resistance from speakers defending human understanding and control. https://www.latent.space/p/aiewf-daily-dispatch-agency
John D. Villarreal, JD, MBA, MA
I'm now a Proficient Builder on @Replit!
Level 3 @amasad - let's GO!!!
shreya rajpal
Re @swyx with the 🐐 takes on making a talk
rejigged my whole talk around the thesis slide for AIE world fair
swyx @aiDotEngineer WF: lots of folks prepping talks next week (congrats!). Some thoughts from RLing on thousands of hours of engineer- and researcher- focused talks:
- AI generated svgs > AI generated imgs. MAXIMUM 4 ai slop images in your slides, I don't care how pretty your mom thinks they are
Shann³
how gbrain's dream cycle works
most knowledge bases degrade as you dump more things in, they go stale, duplicates pile up, citations break, and after a week or two you stop using it
gbrain's fix for that is the dream cycle, a 24/7 loop that works on the brain on its own, mostly overnight
every dream cycle it:
> ingests the day, the meetings, emails, notes and calls, and files them into the brain
> enriches every person and company it saw, adding new details so their pages stay current
> consolidates the day's conversations, turning raw chat into clean, synthesized pages
> merges duplicate pages, spotting two pages about the same thing and combining them into one
> fixes broken citations, repairing links and sources that no longer point anywhere
> ranks each page by importance, so the ones you reference most surface first in a search
> flags pages that contradict each other, so you catch the conflict instead of trusting the wrong one
> preps tomorrow, lining up the next day's tasks and context
Shann³: how embeddings make Gbrain smarter than a plain LLM wiki
a plain LLM wiki finds pages by matching the words you typed. if the wording doesn't match, it misses the right page. gbrain adds embedding models, and that's the real upgrade
think of it as a map of meaning
an embedding
Sai
This is generated from @Replit animation. I am completely impressed by the level of localisation and the quality of the render.
Maximillian Piras
Had a great time at @aiDotEngineer this year, unfortunately not nearly enough as I had to leave early for some family matters. It's amazing to see the community built by @swyx & co keep growing / evolving. Will have many talks to catch up on once they're posted!
AI Engineer: Three days in one building and it still doesn't feel like enough. Day 3 of World's Fair.
sam lessin 🏴☠️
We don’t take stakes in healthy companies. Proposing that as last ditch regulatory capture attempt in commodity market is 😢 … or at best un-American disingenuous WWE politics, not pro-market, not pro-freedom.
Please oppose ACA-7
If you live in California take a moment to click the link here and tell your representatives
We want an end to racism of all kinds in our schools
Tony Guan: Racial preferences in higher education are back on the table with ACA-7, and this could lead to a slippery slope where the exception overwhelms the rule. As Gail Heriot, a law professor and member of the U.S. Commission on Civil Rights, warns, this move could undermine the
I think the days of companies relying only on frontier models are coming to an end.
1. Tokenmaxxing at frontier-model API prices makes no sense.
Uber burned through its entire 2026 AI budget in 4 months, Microsoft moved engineers off Claude Code due to cost, and companies are realizing that running everything on frontier models can get expensive fast. Tokenmaxxing makes sense when you’re on a subsidized $200/month plan, but it’s unsustainable at API rates.
2. Companies will rely on a portfolio of models.
Coinbase recently cut its AI spend nearly in half by switching engineers to Chinese open-source models like GLM and Kimi. Airbnb and Pinterest have done the same with Alibaba’s Qwen models. I believe this will be the default path forward: using frontier models for high-stakes work and cheaper models for everything else.
3. China’s open-source strategy is working.
Chinese models are taking market share from frontier models at US companies. China is also building the full AI stack, from energy, like solar and nuclear, to data centers to domestic chips. The Chinese government is planning a $295B investment in AI data centers, with at least 80% of chips built domestically.
4. Frontier labs are in a Catch-22 situation.
If they release great open-source models, they might undercut their own frontier API revenue. If they gate the best models behind a trusted list, companies will just lean on open alternatives more. The last major US open-source model was OpenAI’s gpt-oss series back in August 2025, which already feels like decades ago in the AI space.
5. The US needs to think about its AI strategy holistically.
I believe restricting access to frontier models will only hurt American innovation. Banning US companies from using Chinese models won’t work either. Just look at how China took over the global electric vehicle market. To maintain our edge, we need the best closed and open models while scaling our energy grid and data centers much faster.
📌 More in my recent essay: https://creatoreconomy.so/p/18-hot-takes-on-where-ai-is-headed
Peter Yang: Here's my new post with 18 hot takes on where AI is headed next:
1. The frontier-only AI stack is collapsing
Companies will use a portfolio of models to save costs, with everyday tasks defaulting to low-cost open models from China.
2. The AI super app era is here
Codex,
Peter Yang
I think the days of companies relying only on frontier models are coming to an end.
1. Tokenmaxxing at frontier-model API prices makes no sense.
Uber burned through its entire 2026 AI budget in 4 months, Microsoft moved engineers off Claude Code due to cost, and companies are realizing that running everything on frontier models can get expensive fast. Tokenmaxxing makes sense when you’re on a subsidized $200/month plan, but it’s unsustainable at API rates.
2. Companies will rely on a portfolio of models.
Coinbase recently cut its AI spend nearly in half by switching engineers to Chinese open-source models like GLM and Kimi. Airbnb and Pinterest have done the same with Alibaba’s Qwen models. I believe this will be the default path forward: using frontier models for high-stakes work and cheaper models for everything else.
3. China’s open-source strategy is working.
Chinese models are taking market share from frontier models at US companies. China is also building the full AI stack, from energy, like solar and nuclear, to data centers to domestic chips. The Chinese government is planning a $295B investment in AI data centers, with at least 80% of chips built domestically.
4. Frontier labs are in a Catch-22 situation.
If they release great open-source models, they might undercut their own frontier API revenue. If they gate the best models behind a trusted list, companies will just lean on open alternatives more. The last major US open-source model was OpenAI’s gpt-oss series back in August 2025, which already feels like decades ago in the AI space.
5. The US needs to think about its AI strategy holistically.
I believe restricting access to frontier models will only hurt American innovation. Banning US companies from using Chinese models won’t work either. Just look at how China took over the global electric vehicle market. To maintain our edge, we need the best closed and open models while scaling our energy grid and data centers much faster.
📌 More in my recent essay: https://creatoreconomy.so/p/18-hot-takes-on-where-ai-is-headed
Peter Yang: Here's my new post with 18 hot takes on where AI is headed next:
1. The frontier-only AI stack is collapsing
Companies will use a portfolio of models to save costs, with everyday tasks defaulting to low-cost open models from China.
2. The AI super app era is here
Codex,
Lukas Kuhn
🔥 We introduce LeVLJEPA: the first fully non-contrastive end-to-end vision-language pretraining method competitive with CLIP & SigLIP 💪🏼
👀 No negatives. No temperature. No momentum encoder. No teacher-student.
TL;DR: LeVLJEPA learns image to text structure by prediction: each modality predicts the other's embedding, while SIGReg keeps each embedding isotropic Gaussian. 🧵
📄 https://arxiv.org/abs/2607.00784
Abhishek Bhardwaj
Thank you @swyx - last year you gave me an opportunity when I was alone building in a corner and this year you backed me with a 45-min slot. These talks push me to make deep OS and infra content accessible to all. Absolutely humbled by the attendance.
swyx @aiDotEngineer WF: for what it's worth, i only invite double-length track keynotes when I'm very sure that both speaker and content deserve it. Today, @chrmanning and @abshkbh did double duty at AIE and by all accounts* people loved the opportunity to go deeper on sandboxing and world models. Look
nxthompson
The last paragraph of @AnneApplebaum’s piece on what patriotism means ahead of our 250th Independence Day. https://www.theatlantic.com/ideas/2026/07/america-250-july-4-idea/687749/?utm_source=twitter&utm_campaign=ntatl&utm_medium=social
Brian Anderson
The Democratic Socialists of America are increasingly a vehicle for hardcore Marxists.
At the DSA’s 2025 convention, as the indefatigable Stu Smith reports, a “coalition of Communist caucuses” seized a majority on the party’s National Political Committee. They repealed the existing ban on Leninist “democratic centralism,” backed a Maoist with a soft spot for revolutionary bloodletting on a security commission, and narrowly passed an amendment calling to get rid of the US presidency and Supreme Court for bodies subordinate to Congress.
Media outlets scoff at the “Communist” label, but the facts are there: self-described Marxists, Marxist-Leninists, and Maoists now have top positions in America’s largest socialist group.
The DSA represents a radical faction with expanding influence in our politics. It’s not a good development, and needs to be stopped.
Read Stu’s full piece: https://www.city-journal.org/article/democratic-socialists-of-america-communism
Randall Balestriero
Oops, SIGReg did it again! Large scale (CC12M->Datacomp-L) vision-language JEPA pretraining beats CLIP and SigLIP objectives! Thanks to SIGReg, our LeVLJEPA has no collapse, no EMA, no stop-gradient, no negatives, no problem! Checkpoints/demo are live: http://levljepa.github.io
Lukas Kuhn: 🔥 We introduce LeVLJEPA: the first fully non-contrastive end-to-end vision-language pretraining method competitive with CLIP & SigLIP 💪🏼
👀 No negatives. No temperature. No momentum encoder. No teacher-student.
TL;DR: LeVLJEPA learns image to text structure by prediction: each
Watching our product owners pitch and demo internally with @v0s feels like science fiction, especially with our design system fidelity. It gives the company end-user empathy and decision-making velocity that we never had with text or slides. Demos > memos.
conor brennan-burke
what surprised me most about yc was how many wildly successful alums come back for their next startup
we had the founder of cruise in our batch
since then the founders of zapier, webflow, firebase, circle medical, etc have all returned
these founders could raise enormous rounds on pedigree alone
i even met someone who had been a yc partner and still came back with a startup
they all had plenty of other options
yet they still chose yc
Michael Sinanian
last day at @aiDotEngineer and i'll be at the Expo's poster area explaining the year's best survey paper on Agent Memory (Hu et al), as we did for @latentspacepod's Paper Club live from the floor with @vibhuuuus. come by and see all the great research (+hot take poasters)! @swyx
Everyone in NYC has the same dog. It's a french bulldog. This needs to be studied
Replit ⠕
Fable 5 is back on Replit!
Especially great for longer, harder projects.
Toggle on High effort mode in Replit Agent and try it today on your toughest builds!
Private connectivity between services is 🔑 for backends.
1️⃣ Register 𝚋𝚒𝚗𝚍𝚒𝚗𝚐𝚜 in 𝚟𝚎𝚛𝚌𝚎𝚕.𝚓𝚜𝚘𝚗 (1LOC)
2️⃣ Read it:
𝚗𝚎𝚠 𝚄𝚁𝙻("/𝚊𝚙𝚒", 𝚙𝚛𝚘𝚌𝚎𝚜𝚜.𝚎𝚗𝚟.𝙱𝙰𝙲𝙺𝙴𝙽𝙳_𝙸𝙽𝚃𝙴𝚁𝙽𝙰𝙻_𝚄𝚁𝙻)
Works for Node, Python, 𝙳𝚘𝚌𝚔𝚎𝚛𝚏𝚒𝚕𝚎… you name it
Vercel Developers: You can now securely connect apps on Vercel using internal URLs.
Learn how to configure Service Bindings ↓https://vercel.com/changelog/secure-internal-communication-between-services
Intelligence is now metered 🥲
Have to use this very wisely
A conversation with Boris Cherny and Cat Wu on the path from Claude Code to Claude Tag, and how it spread from engineering to the rest of Anthropic.
Claude Fable 5 is now available in Claude Tag.
Jenna Pederson
Talking about mental health and managing burnout in the world of 996 at @aiDotEngineer with @swyx and @mikeyk:
“There is no job so important that you can’t be offline for a couple of days.”
👏👏👏👏👏👏👏👏
sam lessin 🏴☠️
The Narrative Strategy of Sam Altman on CNBC today…
Some notes from discussing this ‘we will give America 5% of OpenAI’ idea…
1 - Strategically it makes sense for OpenAI to Invite This
Matrix-multiplication / inference is a commodity… but OpenAI is actually in a WORSE place than utilities which are regulated / have regulatory capture AND there are geographic realities / limits on shipping energy. The only way for an OpenAI to ‘win’ / generate long-term profit is regulatory capture, so why not invite it… this has never been more apparent with the open-source global models coming online
2 - OpenAI has the least to lose / most to gain…competitively smart PR move
They are now behind Anthropic in the ‘pure play’ race… and unlike Google and Meta don’t have actually profitable real business… so a 5% dilution to them is nothing really…. It is worth 5% to hurt their competitors if they can drive the meme.
3 - The deal for America sounds way better than any practical reality.
I actually like the ‘trump accounts’ / deal Americans broadly into the market … but 5%… that is like $100 bucks per person, maybe $150 at a 1T valuation… and these ‘dividends’… of what? A non-profitable company — even if you buy ‘massive growth story…. This isn’t needle moving in any practical way, it just sounds good and is something Bernie and Trump can weirdly both like for different reasons….
Upshot — good on Sam Altman as a cynical political strategist - makes total sense…. Drives AI exceptionalism story, ‘so dangerous’, etc.. in a politically advantaged way…
But on the eve of America’s 250th Birthday … shame. Deeply anti-American, anti-Freedom, anti-Capitalist… and highly dangerous behavior to anyone who actually believes in American values…. And it is exactly the type of outcome starting with bank bailouts everyone worried about in 2008** don’t give mice cookies**
I am interviewing @NousResearch Hermes co-founder @karan4d tomorrow, what topics would you like to see us cover?
I already covered setup, integrations, and basic cron jobs in a previous tutorial - so what else would y'all like to hear about?
This chart is a vertical line lol
Garry's List
According to a new Civic Grand Jury report, 25% of drug OD deaths in San Francisco happened in taxpayer-funded permanent supportive housing.
In one case, it took almost two weeks for staff to discover that a tenant had died — despite promised check-ins every 72 hours.
We've been explaining AI Gateway as a C̶o̶n̶t̶e̶n̶t̶ Token Delivery Network. Like a CDN for AI models.
One great feature of CDNs is the ability to dynamically re-route or deny traffic without redeployment.
When Fable was suddenly retired, we worried about production workloads depending on it. Even Fable aside, models get retired quite often as GPU capacity is heavily contested. Yet our data shows people still having happy production traffic of older model versions!
We're solving this now with AI Gateway Rules. e.g.: you can now rewrite model "routes" on the fly!
𝚟𝚎𝚛𝚌𝚎𝚕 𝚊𝚒-𝚐𝚊𝚝𝚎𝚠𝚊𝚢 𝚛𝚞𝚕𝚎𝚜 𝚊𝚍𝚍 \
--𝚝𝚢𝚙𝚎 𝚛𝚎𝚠𝚛𝚒𝚝𝚎 \
--𝚜𝚘𝚞𝚛𝚌𝚎 𝚊𝚗𝚝𝚑𝚛𝚘𝚙𝚒𝚌/𝚌𝚕𝚊𝚞𝚍𝚎-𝚏𝚊𝚋𝚕𝚎-𝟻 \
--𝚍𝚎𝚜𝚝𝚒𝚗𝚊𝚝𝚒𝚘𝚗 𝚊𝚗𝚝𝚑𝚛𝚘𝚙𝚒𝚌/𝚌𝚕𝚊𝚞𝚍𝚎-𝚘𝚙𝚞𝚜-𝟻
This adds to the repertoire of capabilities and techniques AI Gateway brings to the table to recover lost tokens. If you drop tokens, you lose revenue and customers 💸
Kudos to @rtaneja_ @crowprose @shaper – we went from CLI design to production-grade capability very quickly!
Vercel Developers: AI Gateway now supports routing rules.
Migrate to new models by setting rules in the CLI, no code changes in your app or agent required. https://vercel.com/changelog/ai-gateway-routing-rules
Gabriel Jarrosson
The guy who built FaZe Clan ($2B at peak!) is now building in YC…
Yousef Abdelfattah just got into YC's summer batch with a company called TryNearby. It automates word-of-mouth marketing for restaurants and local businesses by matching them with vetted local creators, so a place that can't afford a marketing team gets a steady drip of social content from people who actually live nearby.
Bullish that it's him building it and not somebody else…
He co-founded FaZe Clan when he was 14. It started as a Call of Duty montage channel and became one of the biggest entertainment brands on the planet, went public in 2021 at $750 million, and crossed $2 billion within months.
He spent his entire teenage years and twenties learning, at the highest level anyone has, how content turns into attention and attention turns into a brand.
So TryNearby isn't some random YC idea he landed on - he's been an expert in since he was 14, pointed at small businesses instead of gaming.
Every restaurant owner knows word-of-mouth is the best marketing they have and nobody can figure out how to manufacture it on purpose. He spent fifteen years manufacturing it on purpose, just at a scale most people will never touch.
That's pretty solid founder-market fit… A person who has done the exact thing the company does, at a level almost nobody else on earth has, now building it as software for people who'll never have his instincts.
The FaZe part is a fun headline. But investors are paying attention as he spent 15 years figuring out how to make people talk about something online…. Now he's packaging it for restaurants, who need exactly that and have no idea how to do it.
I havea bunch of YouTube adsense revenue trapped in my account b/c Google won't confirm my tax info...and there's no way to reach a real human to talk to.
Anyone have any connections who work at YouTube or Adsense?
Gergely Orosz
Re Part 2, w @ankrgyl, @bernhardsson, @Sirupsen, @jerryjliu0, @travers00, @Steve_Yegge, @addyosmani
And what a game @grinich
And many others! Also @swyx AIE is 🔥, how is it better every year??
very proud that the biggest applause line in the AIE keynotes this year was normalizing men talking about their feelings and mental health in hypergrowth
thanks @mikeyk for indulging all my cheeky questions on Fable and Tag!!
Jenna Pederson: Talking about mental health and managing burnout in the world of 996 at @aiDotEngineer with @swyx and @mikeyk:
“There is no job so important that you can’t be offline for a couple of days.”
👏👏👏👏👏👏👏👏
Thought I’d share a fun activity I did with my 8-year-old daughter and Codex today:
1. She drew a picture of a dragon.
2. We uploaded her drawing to Codex and used image gen to make a few more dragon poses. She used voice to give Codex feedback until we got the result below.
3. We sent it to customstickers(dot)com to print sticker sheets (it's about $20 for 10 sheets).
Now she can share her stickers with her friends when they arrive in the mail next week. Try it if you have kids stuck at home this summer :)
Peter Yang
Thought I’d share a fun activity I did with my 8-year-old daughter and Codex today:
1. She drew a picture of a dragon.
2. We uploaded her drawing to Codex and used image gen to make a few more dragon poses. She used voice to give Codex feedback until we got the result below.
3. We sent it to customstickers(dot)com to print sticker sheets (it's about $20 for 10 sheets).
Now she can share her stickers with her friends when they arrive in the mail next week. Try it if you have kids stuck at home this summer :)
everyone come down to the AIE Expo at 12.30, there's a SPECIAL **FLASH** SURPRISE (you'll know it when u see it) and then come to Expo Stage 2 for a very special @latentspacepod live pod with Etched!
Jessie Frazelle: @MartinShkreli But where’s the hot chips talk level of detail
Lynsey Smith
Okay @greptile I was unfamiliar with your game
@aiDotEngineer
💥Susan Dyer Reynolds🗞️
They’re paid agitators. It’s everywhere.
💥Susan Dyer Reynolds🗞️: Why are all the “Trans comrades” wearing masks and trying to block @war24182236 from his right to film in public?@Scott_Wiener perhaps you should ban them from wearing masks like you want to do to law enforcement since wherever they go vandalism follows.
going to world cup games is always awesome, but watching the USA win in the USA during USA birthday week was just incredible
T Wolf 🌁
It's actually closer to 30% of all OD deaths in supportive housing. Plus, if you add in those who died of cardiac arrest etc. after years and years of hard drug use, it's even higher. This is not how we "solve" homelessness. @garrytan
Garry's List: According to a new Civic Grand Jury report, 25% of drug OD deaths in San Francisco happened in taxpayer-funded permanent supportive housing.
In one case, it took almost two weeks for staff to discover that a tenant had died — despite promised check-ins every 72 hours.
John Lindquist
Massive *THANK YOU* to @swyx and the @aiDotEngineer crew for the wonderful birthday surprise yesterday. So super thoughtful. And thanks to all the attendees for singing so beautifully :)
(My wife kept it a secret so she knew to film it. I had no idea what was happening)
Elena
picked up this pin from the ‘72 election at the military surplus store (for a 4th of july costume) and it got me thinking that the nixon v mcgovern election was the greatest refutation of nominative determinism in history. one the one hand you have george mcgovern, a guy who’s name is literally “founding father mcborntorule” against a guy whose name translates to “dick nullificationoffuturesons.” and yet the son-nullifier wins in a landslide. really makes you think.
Alex Cheema
Full house for Local AI Track @aiDotEngineer
We’re going to make Local AI The Default
Stacked lineup:
@nvidia x @exolabs x @OsmanticAI x @roboflow x @PrimeIntellect x @MatthewBerman x @arcee_ai x @huggingface x @UnslothAI x @ollama x @OpenRouter x @cognition
Have you no imagination?
Do you understand what this being possible with one prompt means?
Mark Shust: appendix C, sorry but flying through Hogwarts with no actions is like the most boring game ever: https://x.com/mattshumer_/status/2072511611705692298?s=20
adlin is bts @ ai engineer wf
The star @mochipomsky is diligently listening to her favourite person @swyx
Mada Seghete
What an honor to curate the first AI in GTM track at @aiDotEngineer 😆 Heard that we need a bigger room next year @swyx 😊😅
Gregor Zunic
Most powerful browser agent yet
Browser Use: Introducing: Browser Use CLI 3.0 🌐
Turn any model into a SOTA browser agent
> Direct CDP control via browser-harness
> Run on cloud browsers or real Chrome
> 6× smaller, fewer tokens
Try it now 🔗↓
Fred Wellman
China added over 543 Gigawatts to their grid last year. 434 GW was renewables because they are faster to build. The U.S. added a paltry 53 GW. We are losing the “AI race” to China because these idiots hate green energy. This is suicide by stupid.
Secretary Chris Wright: I'm thrilled to report that after 35 years, on July 4th, we will end the subsidies for new wind and solar projects, thanks President Trump’s Working Families Tax Cut.
Mike Bradley
“Within the next 18 months, you will be able to host GLM 5.2 equivalent intelligence on an RTX 5090 GPU.”
-Ahmad Osman, AI World’s Fair
Kenny D
I just left the final day of the @aiDotEngineer World's Fair Conference in San Francisco. Kudos to @swyx for putting together a world-class lineup of speakers and workshops! It really was an invigorating 4 days.
A few reflections:
-> The open-source AI developer experience is so much more exciting than enterprise
Sure, the expo floor had its polished, enterprise-ready booths. But the energy was in the talks — open-source projects free to build however they see fit.
Highlights:
@pbakaus - Truly on the cutting edge of AI development. Anyone doing UI work should check out his project @impeccable_ai. What he's built is pure wizardry
@CoreyGallon - web automation extraordinaire. I thought I was pretty good with playwright but his claude-agent solution blasted those perceptions aside. Jaws were on the floor
@BHolmesDev - super engaging and informative speaker, showing off Hubble and the open source @warpdotdev
-> We're so early
They say startups are like building a plane while you're flying it, but this seems to be true for the AI industry as a whole.
Memory is still in its infancy, maybe because the best memory systems need to be custom tailored to a specific environment or use case. @RLanceMartin touched on this in his talk, and I'm looking forward to seeing his posts on dreaming
Local AI is barely out of the gate, too. Consensus seems to be that running your own setup is still not for the faint of heart, even for technically-minded folks, though this appears to be changing soon. @TheAhmadOsman is a legend for all of his work distilling the complexity
-> Velocity is up. Code quality isn't.
"Code is provable. Software is not." was the line from @tariqshaukat's talk on verifiability: AI speeds up development but multiplies bugs and maintenance along the way. Models will keep getting better at writing working code — but will they ever match tier-1 code that's perfectly adapted to a specific business?
Either way, I'm leaving inspired, with a long list of things to tackle. Happy building, everyone 👋
Andy Konwinski
http://x.com/i/article/2072403747238723584
Jeremy Daly
“Half of the companies here at @aiDotEngineer could be a markdown file!” ~ @theo 🌶️🌶️🌶️
Jerry Liu
3 years ago I gave a talk at the first @aiDotEngineer conference on "Advanced RAG" techniques in order to work around the limitations of naive RAG.
It's insane how much the world has changed since then, and the world has evolved into standardized, higher-level abstractions around agent harnesses and context.
Some general patterns:
1. Retrieval complexity can be encoded at the agent layer. This means that you can give relatively simple but performant search tools to an agent (e.g. really fast bm25, vector search), and let the agent reasoning enter the right queries to find the right results.
2. To some extent this is still evolving, but I do think we will increasingly care less about "hacking" the context window and more about deciding what business context is relevant in the first place.
3. The way we build agents has fundamentally changed from defining code, to defining runbooks, to defining goals.
Big congrats to @swyx and the entire AI Engineer team for continuing to put out awesome conferences every year.
