The anti-data center stuff is insane and clear evidence that we have entered a post-truth era where many people can be convinced, often by malicious actors, to abandon reality and allow conspiracies/falsehoods to drive policy preferences.

AI helped create a cancer vaccine. Data centers are lowering people’s property taxes, revitalizing dying towns by flushing them with revenue, and…helping power the infrastructure to cure cancer.
Every data center hater should explain why any costs outweigh those benefits.
The Wall Street Journal: An experimental cancer vaccine kept melanoma from coming back. It’s the first successful late-stage mRNA study validating decades of custom-treatment research https://on.wsj.com/4qy7Ule

talked to a YC founder today who pivoted recently but is already at $25k/month, with one GTM trick (super sneaky)
all he's doing is scraping linkedin for the phrase "anyone got recommendations for" + emailing the posters within 2 hours
and yes, i know what you're thinking
"there's no way there's enough posts like this"
there's not, BUT it pulls up similar posts, and he scrapes the people who LIKED those posts
here's why it works:
someone who types "anyone got recommendations for a CRM" on linkedin just publicly declared their buying intent with a timestamp on it
that's not a lead you find in a database
that's a lead that found you first
here's the exact phrases worth searching every morning:
"anyone got recommendations for [tool]"
"alternative to [competitor]"
"looking for a good [software]"
"got burned by [company], where do i go now"
"has anyone used [company], are they worth it"
every one of these is a public declaration of buying intent with a timestamp on it
someone posted this on linkedin 3 hours ago:
"we just outgrew our CRM. need something that can scale with us. open to recommendations"
saw it at 9am
pulled the poster's profile
found the verified email
sent this at 10am:
"hey [name], saw your post about outgrowing your CRM. we help teams in that exact transition, last client we moved over in 2 weeks and cut their admin time by 40%. free 15 min call?"
reply at 10:14am
call at 2pm
signed by 4pm
7 hours from public linkedin post to signed deal
the playbook:
step 1: identify the keywords your buyers use when they're in pain ("outgrowing," "replacing," "alternatives to," "burned by," "need recommendations for")
step 2: search linkedin for those exact phrases
step 3: filter for recency. 3 hours old beats 3 days old by a lot
step 4: also scrape the people who LIKED similar posts, they have the same problem and nobody is emailing them
step 5: enrich the poster, verified email, send 2 lines referencing exactly what they wrote
step 6: sequence the reply within 30 minutes of a positive response
the reason this works:
the declaration IS the qualifier
you don't need to discover their pain
they published it
the sales conversation is already 90% done before you say a word
and you can use Origami AI which searches linkedin by keyword in real time
same prompt you'd type into google
returns people actively discussing the problem your product solves
with verified contact info attached
someone who just posted "outgrowing our CRM" is a better lead than anyone who fits a firmographic profile
because intent beats fit every time
everyone is fighting over the same 210M contacts in apollo
the buyers typing their problems out loud on linkedin are completely uncontested
go find one before your coffee gets cold

Moderna/Merck just ran a 1,137-patient Phase 3 trial where every single dose was unique to that patient's tumor. It worked.
The pipeline: surgical resection → whole exome + RNA sequencing → ML neoantigen ranking → mRNA encoding up to 34 patient-specific targets → manufactured and shipped in 8 weeks. One drug, different sequence for every patient.
The ML step is worth to note: the algorithm ingests WES + RNA-seq to identify somatic mutations, then predicts which of those will actually be immunogenic, ie, displayed on tumor cell surface and trigger a T-cell response. It's designed to keep learning from accumulated clinical and immunogenicity data across patients, not just per-patient.
INTerpath-001 (Stage IIB-IV resected melanoma, 2:1 randomized): combination with pembrolizumab beat Keytruda alone on both primary (RFS) and key secondary (DMFS) at interim. Phase 2b at ASCO 2026 showed 49% reduction in recurrence/death, 59% in distant metastasis/death at 5 years. Phase 3 confirmed both.
What this validates:
— tumor-specific neoantigen prediction by ML works in a blinded trial at scale
— 8-week personalized mRNA manufacturing is operationally real
— effect is additive on PD-1 blockade, not redundant
This is first positive Ph3 for individualized neoantigen therapy. First positive Ph3 for any mRNA cancer therapeutic.
What a great time to live in!! This is the best time for AI & biotech!
Moderna: Moderna and @Merck today announced positive topline results from the Phase 3 INTerpath-001 trial evaluating adjuvant treatment with intismeran autogene, a novel investigational mRNA-based individualized neoantigen therapy jointly developed by Merck and Moderna, in combination

Cooking up ways to teach YC batches the nature of Product Quality.
iPod, Google Search, Instagram — these are singular products that touched 1B+ people, and their creators mastered the craft but critically:
Knew what to build, and what to delete.

The newer generation cancer vaccines aren't like the HPV shot (that prevents a virus). These are therapeutic, they treat cancer you already have. And 2026 is the year they stopped being sci-fi.
The idea: after surgery, sequence your specific tumor, find the mutations unique to it, and build a custom mRNA vaccine that trains your immune system to hunt any cell carrying them. A wanted poster for your personal cancer.
Where we actually are:
Melanoma: just like in immunotherapy 15 years ago the furthest along. The press release of positive Phase 3 made the headlines yesterday, the first ever for a personalized cancer vaccine. Earlier data: ~49% lower risk of recurrence. Heading to regulators.
Pancreatic cancer: arguably the most stunning. In an early trial, patients whose immune systems responded to the vaccine were far more likely to be alive 6 years later, in a cancer where 5-year survival is barely better than 10 %. Now in a larger trial.
Kidney, lung, bladder, brain: all in trials, mostly early phase.
The honest caveats: these work best after surgery to stop recurrence, not to melt large tumors. They're personalized, so they might cost $100–300k and take weeks to build per patient. And most are still early-phase. First approvals at best 2027.
But the direction is unmistakable. The mRNA platform some parts of the internet spent years calling poison - because they got fooled by bad actors making money on their lies - is turning into the most personalized cancer medicine ever made.
These are the main trials currently running:

A milestone for our infrastructure: our first NVIDIA Vera Rubin racks are here and now running our training stack. This is an important step as we expand compute that powers OpenAI's next generation of frontier AI pre-training.

Tried @Replit Free Mode.
Just insane....
Live link 👇
https://kons-tuck-construction-website--jirobuild.replit.app/
Replit ⠕: Replit Free Mode, powered by @OpenAI GPT-5.6 Luna.
Let’s make intelligence accessible to everyone.

Jacksonville Sheriff presents the data. They've used Flock cameras to solve 5 murders, 17 kidnappings, 6 forcible sex offenses, nearly 100 shootings, and 172 missing-persons cases.
Jax Sheriff's Office: Why do we support the use of license plate readers? Because they make our community safer.
 
Over the past two years, LPRs have proven to be an incredibly effective tool in helping solve crimes, locate missing people, and save lives. Since 2024, they’ve helped solved 15 murders,

Replit just keeps getting better and better...
@amasad👏
Replit ⠕: Replit Free Mode, powered by @OpenAI GPT-5.6 Luna.
Let’s make intelligence accessible to everyone.

3 weeks ago, @runlayer sued Rippling to block our launch of a competing MCP Gateway. They incinerated $ millions of their VC funding, then dropped the whole thing like a hot potato, for nothing (see our filing, below).
So today, we’re dropping our new MCP Gateway. It’s better. 1/

MuScriptor, our music transcription model, can now make sheet music and tabs! We export PDFs and editable MusicXML.
Under the hood, we generate these from MIDI using @musescore + some custom preprocessing.

Musk predicted $2 trillion in savings from DOGE.
The real number ended up being less than $50 billion.
This year, the debt passed 100% of GDP and is on track to break its WWII-era record in the next few years.
📸: @MSNOWNews @Morning_Joe

I grew up in the retirement community my dad ran + saw him search for missing seniors when an unwell spouse got ahold of their car keys.
There are serious convos about public safety tech + Flock cameras, but let’s not forget the lives saved every day
https://nowgeorgia.com/franklin-county-family-credits-flock-with-helping-find-missing-grandmother/

On the sidelines of Paris Startup School, I had one of the most interesting conversations with @james406, the founder of @posthog.
PostHog is going through what feels like a new pivot.
They are known for being the best at product analytics. But now they are using AI to become what James calls a "doing company": finding problems, fixing them and opening PRs while you sleep.
I sat down with James to understand how they got there.

My cancer case is n=1. I'm not holding it up as a model, because most of what I did is not specifically replicable or scaleable. But it is a proof of concept that fully personalized medicine, built around one patient's own biology, is viable in ways it simply wasn't a few years ago. I have lived a version of the future of cancer care that I think is worth aspiring to more broadly.
Many of the tools that made this future possible for me are getting better, faster, cheaper, and more accessible every year. I think what we need next is the infrastructure to deploy these tools, and the community willing to build this infrastructure, one patient at a time.
I want to help that future arrive faster. I wrote an article in The Cancer Letter https://cancerletter.com/guest-editorial/20260814_2/ for the physicians and researchers closest to the patients who could benefit most from this approach, who also believe in this future. If you are working towards building the systems and practices that will make personalized medicine operationally real, we want to collaborate with you too.

To read the curriculum SF will now require every 9th grader to take, a parent had to book an appointment, travel to a district office, and read it under supervision.
What is the district hiding? A required course that recasts Asian American success as complicity in whiteness.

In a system designed to facilitate fraud, the only real crime is exposing the fraud
Paul Graham: How should universities respond to the Arday scandal? Should they crack down on plagiarism and academic fraud? Ghent University's response, apparently, will be to fire the person who reported it.

This was a fun talk!
Why Your Enterprise Tech Stack Isn’t Ready for AI Agents
https://www.youtube.com/watch?v=mav15aW9lLM
AI Engineer: Our AI in Healthcare Track from AI Engineer World's Fair 2026 is now live!
Thesis: what can never be wrong cannot be left to probability.
https://www.youtube.com/watch?v=u6q-byPWUuo&list=PLVeaSLBeOF10
- From Ambient Documentation to Clinical Intelligence: @c_asawa, Abridge
- Guardrails First: Engineering

So @swyx and team are still judging his amazing “Kill My SaaS in one weekend” contest, where you could win $10K to replace his $40K/year conference submission (CFP) and scheduling SaaS.
I think 68 people completed the challenge, and some are downright incredible achievements. It's been amazing working with many of them in the Discord channel for the last 10 days. (The competition deadline was extended by 7 days.)
For me personally, I achieved and exceeded my own goal: we not only killed the SaaS we were using for our CFP process, but as of yesterday, we just killed the adjacent SaaS we used for conference scheduling. And it wasn't out of malice — we actually loved using it for ten years. In fact, whenever I go to a conference and see this tool, I smile and think, "Now here are some organizers that love their attendees."
Two weeks ago, I interviewed Ann Perry, who has been doing our conference operations for 6+ years. The thing she complained about most was creating the outward-facing schedules. It was a multi-hour process, laying out the days, defining the talk slots, lunch, breaks, filling them with talks/speakers, etc.
Worse, after doing all that work, she couldn't clone a day or an event, so she had to start from scratch for every day, and for every event!
Hearing details like this made me realize why we often published schedules so late in the process — I'm guilty making tons of changes to "create the perfect program," each of them creating a costly downstream consequence for Ann. It just wasn't worth it to publish until the schedule stabilized.
Last week, we were under pressure to publish a schedule for our Oct 7-8 — so I wanted to see what the scheduling functionality that was created in my tool (Curtain Call CFP) to conform to Swyx's requirements. I was horrified by what the software factory built...
In my defense, I paid the most attention to the submission review process — it was the part I cared about, because it made me most angry with the tools we used for the past 10+ years. There were many sections that I had the software factory create, designed to pass Swyx's evals, and I didn't pay much attention to them.
When I brought up our scheduling section, my heart sank. I got angry. It was the kind of software that I hate: it presented 24 talk slots, all pre-filled, each requiring the organizer (Ann) to undo it.
This is like the ultimate "eff you" to users — that's 24 fields that someone has to undo. 24 clicks of mindless drudgery and meanness inflicted onto the poor organizer. (See first image.)
So I put into Claude Fable the Ann interview transcripts and the video where it could see the screenshots, where she walked me thru how she has created conference schedules in the past. I told it to read the manifest we wrote for the project (e.g., no one likes filling out forms, no one likes switching pages, etc.) and had it create some design briefs.
Within two hours, we replaced the awful scheduling grid full of hateful dropdown boxes with a sublime Trello-like experience. We could easily change the timeslots merely by changing a duration, and changes would ripple through the rest of the day. We could detect when slots infringed on hard boundaries (e.g., you can't move lunch because of the venue food/beverage staff, you can't change plenary times, because of union AV teams).
We added the ability to add tags to talks, to encode information that we use to make scheduling decisions (e.g., opener/closer, enterprise experience report vs expert/consultant).
In short, we published a production schedule the earliest we've ever done in our history, 20x faster than ever, with the lowest cost of change.
And it's in production use, supporting two of our conferences already.
Some reflections so far:
- technically, the old terrible "make you undo 24 talk slots" that I created for you might have passed the evals -- but it would have been as bad than many of the tools that I've complained about for 10+ years.
Tim O'Reilly famously said, when something becomes free, something else becomes more important — and as creating software becomes free, he predicted what becomes even more valuable is judgement and taste.
- Eval-driven software development is going to change the world, and will allow us to clone SaaS products in a weekend — but the risk is that we recreate the hateful software products that we're replacing. It will be cheaper, but dooms users to the same hell they were stuck in before. (At least we're the warden, I suppose?)
- That 68 people completed Swyx's challenge shows that there are certainly some categories of SaaS that you can replace in one weekend
- There is another danger: great SaaS companies are now able to start moving into adjacent categories and start pushing out incumbents. (A great example from Enterprise AI Summit from April: Tisson Mathew, CEO of Skypoint, describe how they've 10x shipped more products in a quarter than they had in 6 prior years combined -- and they're moving into adjacent categories, easy displacing the incumbent, because they only needed a couple of features.)
You can see our live schedule (created more easily than ever!) for our upcoming Enterprise AI Summit here!
https://curtaincallcfp.com/program/enterprise-ai-summit-charlotte-2026
Images: old screen, new screen, design brief.
Next post will show movie of the Trello-like scheduling experience we've been dreaming of for over a decade.
Gene Kim: So @swyx announced his “Kill My SaaS in one weekend” contest, where you could win $10K to replace his $40K/year SaaS. On Saturday morning, I learned that it was to replace their conference CFP software.
OMG. I've run about 24 conferences over 12 years, but something I rarely

"and I remember you DMd me on twitter"
The value of X (Twitter) is unmatched.
Amjad Masad: This partnership with @OpenAI is long overdue.
Before Replit was in YC, PG asked Sam to recruit us.
Here’s Sam telling the story of how we first met:

In all seriousness I think this represents both a disturbing society-level inability to process basic facts that involve numbers, a lurch toward degrowth on the left and right, and a failure of mainstream journalism to go beyond just nodding along to the "concerns" of everyday people without checking to see if they're the result of a game of telephone that leads back to nothing
Hunter📈🌈📊: Insane decline.

The chances of @CBS picking 5 random Cupertinians and none of them being Asian is 0.28%.
SFGATE: The city of Cupertino is 70.9% Asian. A new TV show set there ignores this. https://www.sfgate.com/sf-culture/article/cupertino-tv-show-22395253.php?taid=6a875cca2f77dd0001a5a572&utm_campaign=trueanthem%2B3984&utm_medium=social&utm_source=twitter

As I was considering joining Replit, I shared a memo about how the cost of consumer vibecoding would go to 0
I wanted to make sure the team agreed before joining. Fortunately, we were on the same page.
We've since scaled up enterprise >10x. And Replit is now free for consumers
Replit ⠕: Replit Free Mode, powered by @OpenAI GPT-5.6 Luna.
Let’s make intelligence accessible to everyone.

Just make great things. Find your moat later.
Nikunj Kothari: the models have no moat (OpenAI, Anthropic, XAI)
the IDEs have no moat (Cursor, Windsurf)
the harnesses have no moat (Cognition, Factory, LangChain)
the app builders have no moat (Replit, Lovable, Bolt)
the wrappers have no moat (Harvey, Abridge, OpenEvidence)
the inference