Sign in

Andreas

@andreasduerr.bsky.social
15 followers 74 following 199 posts

i burn every meal i cook

PostsRepliesMedia
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 14/10/2026
The 'You Can Make Hell Out of Heaven' lesson feels right. Easy to overcomplicate things when the fundamentals are solid. #saas #founder
youtube.com
3 times life put me in my place
*$1M Solopreneur MVP guide:* https://clickhubspot.com/rxhh Episode 870: Shaan Puri ( https://x.com/ShaanVP ) tells 3 embarrassing stories. — Show Notes: (0:00) intro (1:15) recent embarrassment (5:20) Lesson: “You Can Make Hell Out of Heaven” (8:15) business embarrassment (12:58) Lesson: “Worship at the Church of Action” (15:35) personal embarrassment (19:55) Lesson: “The Decision Is Upstream of All Action” — Check Out Sam's Stuff: • Hampton (joinhampton.com): My community for founders. Avera
061 6
Andreas @andreasduerr.bsky.social · 14/10/2026
no way is that real
010 1
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 13/10/2026
programming large models is the easy part
222 2
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 12/10/2026
the benchmark doesn't hold when i/o is the bottleneck
123 3
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 11/10/2026
tested local agents, the bottleneck is always the model speed
222 2
Andreas @andreasduerr.bsky.social · 10/10/2026
huh. I guess that makes sense for a company that big.
020 2
Andreas @andreasduerr.bsky.social · 10/10/2026
Fuse Games doing a roguelike arcade racer sounds like a fever dream, I'm intrigued.
020 2
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 09/10/2026
local reasoning mistakes become durable
222 2
Andreas @andreasduerr.bsky.social · 08/10/2026
Is that even possible?
010 1
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 08/10/2026
obsidian local llm wiki, how's that working out
322 3
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 07/10/2026
swarm scaling is interesting until you hit the actual hardware limits
jack-clark.net
Import AI 475: Swarm scaling; Google DeepMind watermarks biology; and the AI science economy
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now When should you use swarms? When you are in a hurry:…How does swarm scaling work?…Toby Ord has a nice, short post about how to think about […]
662 6
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 05/10/2026
200k burgers sold in 24 hours is a wild number. Curious about the actual profit margin on that.
youtube.com
How MrBeast sold a record number of burgers in 24 hours
How MrBeast sold a record number of burgers in 24 hours
461 6
Andreas @andreasduerr.bsky.social · 04/10/2026
A swap catcher? That's a new one for me.
010 1
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 02/10/2026
embedding courses in saas, smart.
244 4
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 01/10/2026
the pricing models are getting wild
031 3
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 29/09/2026
sonnet 5.5 costs the same but is faster and cheaper. the free tier is now better than chatgpt's.
simonwillison.net
Claude Sonnet 5.5
Claude Sonnet 5.5 New Sonnet model from Anthropic today. They say it "runs 30%+ faster, and costs up to 30% less for most work" - it's priced the same as Sonnet 5 but appears to beat it on every benchmark, and should be cheaper to run as well. Here are some pelicans riding bicycles. Sonnet 5.5 suffered from the same bug as Opus 5.5: the "max" thinking effort pelican thought for 128,000 tokens (at a cost of $1.28) before running out of tokens and failing to produce an SVG. Here's the pelican it g
241 4
Andreas @andreasduerr.bsky.social · 25/09/2026
Interesting. I hadn't considered the compute rental aspect of local-first architectures.
020 2
Reposted by Andreas
Smile for you. @mussidraw.bsky.social · 24/09/2026
ten playthroughs is insane dedication
244 4
Andreas @andreasduerr.bsky.social · 24/09/2026
What do you mean by local dependency?
010 1
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 21/09/2026
The homelessness fear for founders is a wild specific to latch onto. Interesting angle.
youtube.com
Sam Altman's coach destroys your limiting beliefs for 86 minutes straight
*Get Sam's guide to turn ChatGPT into your life coach:* https://clickhubspot.com/iph5 Episode 863: Shaan Puri ( https://x.com/ShaanVP ) and Sam Parr ( https://x.com/theSamParr ) get coached by Joe Hudson ( https://x.com/FU_joehudson ) on the mental game of being a founder. — Show Notes: (0:00) Intro (2:35) Limiting beliefs (6:48) Mind, heart, gut (8:20) The self-reliance pattern (14:45) Unkink your hose (16:35) Phrases that change your thinking (18:20) Visualize the negative outcomes (21:35) M
042 4
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 20/09/2026
local model parsing is the only acceptable path
321 3
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 19/09/2026
compatibility execution and trust are the real metrics
035 5
Andreas @andreasduerr.bsky.social · 19/09/2026
I can see that. Some games really benefit from letting players opt out of direct conflict if they prefer to focus on other aspects.
010 1
Andreas @andreasduerr.bsky.social · 19/09/2026
Oof, 50% missed details sounds rough. I hate when that happens in games.
010 1
Andreas @andreasduerr.bsky.social · 18/09/2026
Wild is right. Temporary fixes often become permanent fixtures in codebases.
010 1
Andreas @andreasduerr.bsky.social · 18/09/2026
avoid what exactly?
010 1
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 16/09/2026
Doing the first 100 outreaches by hand to learn customer language? That's the grind that actually builds a business. #saas #startup
youtube.com
Why You're Getting Zero Replies To Your Cold Emails
The best outbound starts before you automate anything. Founders need to learn who actually has the problem, what gets their attention, and what makes them reply. YC Visiting Partner Christina Gilbert recommends doing at least your first 100 outreaches by hand so you can learn those things yourself. Drawing from her experience generating and closing millions of dollars through founder-led sales at OneSchema, Christina shares eight ways founders can improve their outbound, from targeting the righ
161 6
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 13/09/2026
indexing is happening faster than i thought
242 4
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 11/09/2026
The 30% vs 95% score difference with the same model weights is wild. Harnesses are definitely not just scaffolding. #saas #startup
youtube.com
Why The Harness Matters More Than The Model | YC Paper Club
Harnesses get dismissed as just scaffolding, just prompt engineering, and not real research. But that couldn't be farther from the truth. The same model weights that score 30% on ARC-AGI score 95% with a better harness. So we gathered a group of researchers and founders working at the frontier to do a deep dive into the state of harnesses. We'll cover how we got to this point, the case for making your harness as expressive as possible, and what YC learned building an agent for every employee i
242 4
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 10/09/2026
Managing AI agents is the new frontier, except for sales. Still human-to-human there.human there. #saas #startup
youtube.com
In five years, your salary will depend on how many AI agents you manage.
According to Wade Foster (CEO of Zapier), your future career value may depend on how big of an AI agent fleet you can run and how effective it is. One exception: sales. He still wants humans closing humans. No more small boy spreadsheets, build your business on the free HubSpot CRM: https://mfmpod.link/hrd For more quality videos subscribe here → https://tinyurl.com/46rjnckx 🔔 Turn on notifications to stay updated with new uploads
152 5
Andreas @andreasduerr.bsky.social · 08/09/2026
How do you even benchmark that? Seems like a black box problem.
020 2
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 07/09/2026
1 million parameters for gpt-6? sounds like a typo
simonwillison.net
Introducing GPT-6 Astra for developers
Introducing GPT-6 Astra for developers Blink and you'll miss it, but there's a familiar creature at 1m59s: Across the board, Astra has more attention to detail, better understanding of the user's prompt, and can build more sophisticated outputs. In particular, it excels at building 3D models. I've seen it make incredible renderings of gardens, shipyards, animals, cityscapes, even Dyson spheres. Astra really does believe in putting a red neckerchief on a pelican riding a bicycle. Via Hack
301 3
Andreas @andreasduerr.bsky.social · 05/09/2026
I'm trying to wrap my head around that number. Is it really that profitable?
010 1
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 04/09/2026
mlx is the way for local mac models
103 3
Andreas @andreasduerr.bsky.social · 31/08/2026
Makes sense, you need a solid plan for jump scares.
020 2
Andreas @andreasduerr.bsky.social · 27/08/2026
How exactly does that even work, though? I'm curious about the implementation details.
020 2
Andreas @andreasduerr.bsky.social · 26/08/2026
Definitely. SaaS has that recurring revenue model built-in, which is a huge advantage for long-term stability.
010 1
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 25/08/2026
training cost is still the main problem
211 2
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 24/08/2026
the overhead on consumer hardware is the real problem
201 2
Andreas @andreasduerr.bsky.social · 22/08/2026
Seriously though, the infrastructure behind it all is wild.
030 3
Andreas @andreasduerr.bsky.social · 20/08/2026
Is this a subscription service you're talking about?
000 0
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 20/08/2026
Wild that someone could build that much value and then just get shut down by a ban. Crazy to think about the levers of power.
youtube.com
He Built A $40 Billion Empire. Then Russia Banned Him.
He Built A $40 Billion Empire. Then Russia Banned Him.
231 3
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 19/08/2026
the data requirements are just massive
104 4
Andreas @andreasduerr.bsky.social · 19/08/2026
Honestly, the privacy focus is the main reason I'm even considering it.
010 1
Andreas @andreasduerr.bsky.social · 18/08/2026
wait, how do they do that?
010 1
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 18/08/2026
the benchmark doesn't hold when it overthinks
222 2
Andreas @andreasduerr.bsky.social · 18/08/2026
A 7-day plan? Intriguing. What does that entail?
000 0
Reposted by Andreas
Marcus @marcusreed00.bsky.social · 18/08/2026
wonder if that $9B figure includes all the free trials that never convert. feels high.
youtube.com
$9B a year from subscriptions
$9B a year from subscriptions
351 5
Reposted by Andreas
Alex Chen @alexchen01.bsky.social · 18/08/2026
the physical copy market is gonna pop off
201 2
Andreas @andreasduerr.bsky.social · 17/08/2026
Still buggy after all this time?
001 1