9/ First Substack post on Better Half and how we're capturing what the industry needs 👇
yourbetterhalf.substack.com/p/ai-alignme...
We're raising our pre-seed and hiring founding engineers who want to solve the hardest problem in AI.
DM me.
8/ But that's band-aids on broken reasoning.
We're building tech that's actually good for people. Creating the training data that doesn't exist.
Training models on what alignment actually means — empirically, intrinsically, as a data science problem.
www.better-half.ai
7/ So we had better shape better ones!
@humanetech.bsky.social and researchers like @jonathanhaidt.bsky.social have been sounding the alarm for years on toxicity, addiction, sycophancy, lawsuits from engagement-driven tech.
The response? Regulation, content moderation, guardrails.
6/ When users practice difficult conversations with AI that responds like healthy partners do — with resistance, memory, emotional coherence — they improve at navigating human relationships.
The skills transfer...because first we shape our tools, and thereafter they shape us.
5/ By creating Better Half, a conversation space that models healthy dynamics, encourages your social life, and helps you practice being the partner you're becoming.
Every successful interaction generates training signal on what healthy relational dynamics actually look like.
4/ This becomes the training foundation for safe AI in any high-stakes context — mental health, education, robotics, defense.
AI can't replace human connection. But it can help us get better at being human together.
How?
3/ You can't buy this data. You can't scrape it. You can't generate it.
So we're building the infrastructure to capture what's never been captured:
The first dataset of longitudinal relational dynamics optimized for human thriving.
2/ Social media = engagement maximization
Customer service = utility maximization
Therapy transcripts = pathology, not health
But longitudinal relational dynamics from healthy human interactions?
Doesn't exist...in any dataset.
1/ BIG NEWS!
I launched the venture to solve a problem that's obsessed me for YEARS.
AI alignment is a $200B+ product masquerading as a research question.
Everyone's building "safer" chat apps, but nobody's attacking the underlying issue: training data on relational integrity doesn't exist.
And for broader perspectives from founders and innovators building tech for human agency, we're over at Substack: agentictech.substack.com
If you're not already subscribed to The Agentic Tech Podcast, let's fix that: open.spotify.com/episode/1FGD...
I recently spoke with @deanwb.bsky.social about his post-White House thoughts on AI policy, including the not-at-all-surprising trend in AI safety & alignment emerging as conservative issues.
PSA: The internet didn’t die with Strongbad. Due to redshift of Overton Window physics, overly earnestcore is unironically cool again.
www.tiktok.com/t/ZTYjFAWwy/
Thanks for noticing.
It feels like every 10th post on here mentions blocking and blocklists.
I realize this place represents fresh air for many, but is blocking really how we heal our national discourse? By pretending other people and their opinions simply don’t exist?
Of course! Tethics & Chill, and our upcoming Agentic Tech summit. @chadfowler.bsky.social and I sent you an email!
Why it be this way? And can you draw skiing instead?
For those that made their accounts in 2024 and not before, what brought you here, and which month did you sign up?
@pfrazee.com show us your growth chart. 📈
Apparently the party is officially here on Bluesky. That true?
I'll post episode one as soon as Apple and Spotify let us out of podcast feed submission purgatory! 💀😬
The topic is complex but this episode definitely isn’t.
Elena breaks it down with her characteristic humor, laying the groundwork for the rest of the season by showing why privacy is the pragmatic choice – not just for individuals, but also for democracy, business, and society.
Have I been silent? Guilty.
Have I been busy? So guilty I'm never getting parole.
Peep what we've been working on over at @agency.bsky.social, and give us a follow so you don't miss when the feed drops. For our first episode, Elena Nadolinski of IronFish talks all things ZK.
12/ If this thread resonates, there’s an entire season of podcast episodes like it coming this summer.
Please RT so that others can find out about what we’re building at @agency.bsky.social.
11/ We hope you'll come away with a holistic and interdisciplinary appreciation of privacy's foundational role in defending neural sovereignty, cognitive consent, and agency amid exponential technology.
Everything is still under wraps, but follow @agency.bsky.social to stay informed.
10/ Together with our guests, my co-host
@chadfowler.bsky.social and I approach privacy through a "cross-functional" looking glass.
We explore the intersection of privacy and AI, jurisprudence, finance, regulations, cryptography, philosophy, democracy, and national security.
9/ Our first public artifact, The Privacy Podcast, will be released this summer, exploring why privacy forms the bedrock of a free, healthy, and functional society.
8/ This is a gap I hope want to bridge with the hush-hush work I've been doing at @agency.bsky.social, an off-the-record discourse community of founders, investors, and operators working at the intersection of privacy, crypto, and ethics.
7/ Worse, there are few wise teachers in the space because crypto doesn’t like “boomers” and “boomers” act dismissive and disengaged instead of curious and accountable.
6/ It's galling to see the persistent pull of the panopticon in crypto.
But I understand it:
Crypto builders are 20 years old, have the attention span of horny teenagers, and zero historical memory because they’re, well, 20.
5/ It’s weird to me that people don’t feel alert and mindful when we talk about ID, or worse — get excited about the novelty of negative on-chain reputation.
Yes, please tell me more about taking the worst excesses of web2, but make it immutable.
