Soerensen

6 ms·

1.
After hitting #1 on Product Hunt, ChatGPT became our biggest referral source(brew.new)
3 points by Soerensen 3mo ago|0 comments
2.
by Soerensen 7mo ago
Yes, but also no one seems to care based on the lack of engagement on this post. That's enough for me to draw a definitive conclusion on where the future is headed. Sad, but true.
3.
by Soerensen 7mo ago
Would love to hear your thoughts: https://news.ycombinator.com/item?id=46901199
4.
by Soerensen 7mo ago
The scrappiest thing that worked for me: manual onboarding calls with every single early user, even when it didn't scale. I'd hop on 15-minute calls to understand their workflow, then send them personalized Loom videos showing exa
5.
by Soerensen 8mo ago
Appreciate it! I should clarify that it's not just grammatical. I find that AI can sometimes help me articulate ideas based on my thoughts in ways that I hadn't even considered.
6.
by Soerensen 8mo ago
Nope! https://www.linkedin.com/in/philipsorensen But as a non-native english speaker, I do use AI to help me formulate my thoughts more clearly. Maybe this is off putting? :)
7.
by Soerensen 8mo ago
The approach of building for one specific user (your wife) rather than abstracting too early is underrated. You end up with something that actually fits the workflows instead of a generic tool that needs heavy configuration. Elixir + Ash is
8.
by Soerensen 8mo ago
The WinGUp updater compromise is a textbook example of why update mechanisms are such high-value targets. Attackers get code execution on machines that specifically trust the update channel. What's concerning is the 6-month window. Sup
9.
by Soerensen 8mo ago
The $10 deposit validation approach before committing to manufacturing is underrated. So many hardware projects fail because founders fall in love with the build before confirming anyone will pay. What stood out to me: the factory miscommun
10.
by Soerensen 8mo ago
The "framework fatigue to custom solution" pipeline is well-trodden, and I think it often makes sense for specific use cases. The tradeoff is always: initial development time (custom = longer) vs. maintenance burden (framework = d
11.
by Soerensen 8mo ago
Interesting framing. On your axioms: Axiom 3 (stable global reference frame) seems most practically actionable. In production systems, we've found that grounding the model in external state - whether that's RAG with verified sourc
12.
by Soerensen 8mo ago
Clever application of voice AI. The pain point is real - phone holds are one of those friction taxes everyone pays but no one thinks to solve. A few questions from someone who would use this: 1. How does it handle identity verification? Man
13.
by Soerensen 8mo ago
The positioning shift here is interesting. v0 started as "generate UI from prompts" and is now framing itself as a full app builder with persistent context. What I find compelling about this direction: the bottleneck in AI-assiste
14.
by Soerensen 8mo ago
Definitely possible. I dropped out at 16 and now run a startup after leading growth at Revolut. The key differentiator for non-degree candidates: demonstrated results over credentials. Build something people can see - a side project, open s
15.
by Soerensen 8mo ago
The agent orchestration point from vessenes is interesting - using faster, smaller models for routine tasks while reserving frontier models for complex reasoning. In practice, I've found the economics work like this: 1. Code generation
16.
by Soerensen 8mo ago
The TypeScript + MongoDB combination for AI coding is a smart architectural choice. I've found that schema-less databases reduce the class of errors agents struggle with most - the migration/schema drift issues that require unders
17.
by Soerensen 8mo ago
The observation about agents not using skills without being explicitly asked resonates. In practice, I've found success treating skills as explicit "workflows" rather than background context. The pattern that works: skills th
18.
by Soerensen 8mo ago
The bias-variance framing here maps well to what I've observed building AI-assisted workflows. In practice, systematic misalignment (bias) is relatively easy to fix - you identify the pattern and add it to your prompt/context. &qu
19.
by Soerensen 8mo ago
The "stale symbol" detection is a nice touch - one of the failure modes I've noticed with AI coding agents is them operating on outdated mental models when files have changed mid-session. Color-coding what's potentially
20.
by Soerensen 8mo ago
The induction/deduction/abduction trichotomy is useful, but I wonder if the boundary is as clean as the paper suggests. When Claude or GPT-4 are asked to "explain why X might happen" given sparse data, they often produce
21.
by Soerensen 8mo ago
The JuiceFS + copy-on-write snapshot approach is clever - being able to restore full workspace state (including Docker images and mise-installed tools) to any previous turn is something the commercial cloud agents don't offer. The warm
22.
by Soerensen 8mo ago
The "dev mode" where you can change lambda code live is the killer feature here. The deploy-wait-test-repeat cycle is what makes serverless development so frustrating compared to local Flask/FastAPI development. I see others
23.
by Soerensen 8mo ago
The satire is great, but this actually points to a real gap in agentic architectures. Most production AI systems eventually hit decisions that need human judgment - not because the LLM lacks capability, but because the consequences require
24.
by Soerensen 8mo ago
Our biggest cost multiplier was "conversational drift" - not the initial call, but what happens when you let users iterate. In our email marketing tool, a user might say "make it more punchy" → AI rewrites → "actual
25.
by Soerensen 8mo ago
The distinction is increasingly about taste and judgment, not syntax. When I work with AI tools for building products, the hard part is never "can AI write this function" - it almost always can. The hard part is knowing what to bu
26.
by Soerensen 8mo ago
The burnout section resonated. The pattern of "I will learn everything myself" followed by "I cannot do this alone" is so common among technical founders, myself included. Two things stood out: 1. Your Meta Ads experienc
27.
by Soerensen 8mo ago
The "talk to your book" feature is genuinely clever - having the context window aware of where you are in the text prevents the usual "wait, you just spoiled the end" problem with most PDF chat tools. Question: how are y
28.
by Soerensen 8mo ago
I think you are asking the wrong question. AI has not replaced Excel, but it has replaced the need for me to hire developers to build features. I am a non-technical founder (background in growth/marketing). Two years ago, building a pr
29.
by Soerensen 8mo ago
The honest answer: 90% of marketing is consistency, not brilliance. What's worked for me: 1. Be genuinely helpful first. Before you promote anything, spend a month just answering questions in communities where your users hang out. Not
30.
by Soerensen 8mo ago
I'm a non-technical founder who built an entire SaaS frontend using v0 and Cursor over the past year. My experience might be useful here since I'm on the extreme end of "vibe coding" - I genuinely don't understand m

More ›