Shaun McAvinney
@Smcavinney

Induction.ai Previously @dagster and @stitch_data

 Philadelphia PA
induction.ai
Joined June 2008
201 Photos and videos
axhiii
@axhiii88547
Sep 25
Be honest: what was your first code editor?
2
107
The ending was always my favorite
Once Upon A Time On The Internet@newgroundsdoc
Sep 1
23 YEARS AGO TODAY, @TheWeebl posted BADGERS to @Newgrounds.
1
4
108
People are accepting LI invites today - summer vacation is over.
2
58
The days before school starts are so high stress. Not for me anymore, but the tensions are high in the house.
1
58
Who's the marketing designer I should work with? Claude design ain't cutting it.
2
3
218
Never change a model midstream
Tibo
@thsottiaux
Aug 15
Replying to @albfresco
Because it’s set as an instruction at the start of the context window
3
154
This is why I turn down all podcast requests.
The Hollywood Reporter
@THR
Aug 19
When Seth Rudetsky received an invitation to appear on Nikki Glaser’s podcast, he had no particular reason to think it wasn’t real. Not long afterward, another invitation arrived, this time supposedly from Amy Poehler’s podcast. Rudetsky actually knows Poehler, so this time he employed the most reliable authentication technology available: He asked Amy Poehler. No, she replied, she hadn’t invited him onto her podcast. Rudetsky had stumbled into an increasingly sophisticated scam in which fraudsters impersonate famous podcasters and their representatives, using flattering, highly personalized invitations — most likely composed by AI — to lure prospective guests into handing over money, login credentials or control of social media accounts. hollywoodreporter.com/news/g…
1
78
Is it crunch time? Feels like crunch time.
93
Don't hate the player, hate the game
This tweet is unavailable
81
Everyone is building AI coworkers, what about an AI work friend?
1
1
121
Clay would be great if you didn't accidentally almost bankrupt your company fiddling around.
1
137
Shaun McAvinney retweeted
Pasha Sadri
@pashasadri
Aug 10
I keep seeing AI teams solve the rising inference cost problem by making the product worse. The team compacts context, pushes more work to less capable models, or caps usage. The bill comes down. So does the product. Conventional context compaction can discard details the agent still needs. This is a hard problem to solve.
1
6
107
I'm out of the office, so my agents might reply slower than usual.
2
111
Shaun McAvinney retweeted
Jake Stein@jakestein
Aug 6
We just published the latest version of our SaaS Contract Benchmark report. Link to the full report in the comment below, and here are a few highlights:
1
2
5
91
Shaun McAvinney retweeted
Pasha Sadri
@pashasadri
Aug 5
Long-horizon agents don’t just use more tokens as they work longer. Their input cost can compound quadratically with every added step.
1
3
45
Today's Mission on Tiktok has brought so much joy to our house.
74
I'm surprised the models don't mention bus factor more often.
1
1
81
Shaun McAvinney retweeted
Jeff Barg
@jeffbarg
Aug 4
Claude, rewrite this codebase in Rust. Make exactly one mistake. Don’t tell me what it is, make me search for it. I want to feel something.
6
3
36
4,157
Shaun McAvinney retweeted
clara the urbanist
@SlopHq
Aug 3
draftkings did $6 billion in revenue in 2025 with 4.8 million active users. that is roughly $1,250 per person per year, extracted, from a group that mostly told pollsters they cannot afford groceries.
20
65
984
31,242
Tibo
@thsottiaux
Aug 4
Given some of the results I'm seeing recently, it's pretty clear Codex is a good harness. But it will seem primitive in 2-3 months and we're about to go through another major evolution in how we use AI at the frontier. The next generation of models need more than your laptop.
118
Load more