claude vs codex isn't about which is better. it's about workflow fit codex for autonomous background tasks. claude for interactive reasoning sessions. use both strategically
leanware.co/insights/codex…
English
Paolo Furlan
362 posts
Paolo Furlan
@goreste
Indie app developer and Lead iOS @revolutbusiness
London, England Katılım Aralık 2009
253 Takip Edilen111 Takipçiler
every ai coding assistant claims they understand your entire codebase meanwhile i'm still explaining to claude why that function exists three files over
English
built 6 products in 2025. 5 failed.
the successful one? solved my own problem building the other 5.
meta-problem solutions hit different because you understand both the pain and the context intimately.
English
gabriel bortoleto crashed twice in brazil 2025, almost delayed audi's 2026 car manufacturing.
replenishing 2025 parts meant less time building 2026 components.
one weekend of walls nearly derailed 18 months of f1 planning.
English
portfolio approach beats betting everything on one saas.
pascal levy-garboua: 6 micro-saas → $120k mrr → shut down 1, sold 2, kept 3.
multiple shots, faster learning cycles, reduced existential risk per project.
English
curse of knowledge in ai development:
if you understand transformers, you overestimate what they can do.
if you don't understand them, you underestimate what they can't do.
both groups ship broken features.
English
gemini 3.1 pro: $2 per million tokens
claud opus: $25 per million tokens
the 12.5x price gap means you're not paying for capability anymore. you're paying for convenience and ecosystem lock-in.
English
swiftui is 6 years old and still can't handle basic camera functionality without uikit bridging.
"declarative" and "production ready" apparently don't mean the same thing in cupertino.
English
deepseek v3.2 scores 60% on swe-bench at $0.30 per million tokens.
opus hits 89.4% at $25 per million.
that's a 12.5x price difference. deepseek doesn't need to be 12x worse to win your budget. just needs to be good enough for your use case.
English
indie hacking used to mean "no vc, just build."
now it's "ai built this in 3 days, here's my $50k mrr."
the tools changed. the hustle stayed exactly the same. but nobody talks about the 99 failed experiments before the winner.
English
rork can generate native ios apps using swift, not react native.
literally replaces xcode. test instantly, install in one click.
no more 45-minute compile times for a button color change. this is what happens when ai finally understands mobile development.
English
Everyone's debating Audi's F1 power deficit, but this analysis shows the real story: "Only Ferrari and Audi suppliers have obtained fuel approval with 7-10 days remaining." The sustainable fuel transition might be more chaotic than the engine regulations.
x.com/ThisIsFormu1a1…
This is Formula 1@ThisIsFormu1a1
🚨| 2026 fuels: only Ferrari and Audi suppliers have obtained the OK: — As the 2026 F1 season approaches, fuel regulations are a major focus. Suppliers must now provide gasoline derived from bio-fuel or e-fuel, moving away from fossil fuels. This shift requires new barrels to prevent contamination, complicating the approval process. — The FIA has tasked British company Zemo with overseeing the production and certification of these new fuels. The process is lengthy, and as it's the first year of certification, delays and obstacles are anticipated. Despite this, all teams are expected to compete in Melbourne. — If a supplier fails to secure approval in time, teams will use 'provisional' gasoline. The specifics of this provisional fuel remain unclear, making it difficult to assess any potential disadvantages. Partial approvals allow teams to run with incomplete synthetic gasoline. — Ferrari's supplier, Shell, and Audi's supplier, BP, have successfully obtained fuel approval. Ferrari's SF-26 has already tested the new Shell bio-fuel in Bahrain, setting them apart as the only team opting for bio-fuel, while others have chosen synthetic gasoline. — The regulation changes aim to ensure sustainability in F1, but the transition is not without challenges. Teams have a limited window to secure full approval, with 7-10 days remaining before the season opener. The situation underscores the complexity of adapting to new fuel standards. #scuderiaferrari 🇮🇹 #audif1 🇩🇪 #f12026 VIA: [AutoRacer IT]
English
This comprehensive AI coding comparison reveals something everyone missed: "The 23-point gap isn't the model. It's the agent scaffold doing the work." We're optimizing benchmark scores while ignoring the infrastructure that actually ships code.
builder.io/blog/codex-vs-…
English
70% of micro-saas generate under $1k/month.
only 1-2% exceed $50k/month.
the success stories you read represent the top 2%.
most indies are grinding for years at $500 MRR.
English
2026 saas market: $15.7b → $59.6b projected by 2030.
micro-saas growing 30% annually while big players chase enterprise.
the opportunity moved from competing with salesforce to solving problems salesforce ignores.
English
Paolo Furlan retweetledi
Hey @X . I just made this Mac app to generate your App Store screenshots, it's open source and can be downloaded from here:
github.com/vberihuete/app…
This is a small contribution to the apple developers community!
English
xcode team definitely cancelled swift assist.
they spent 2 years promising ai-powered code completion and shipped... improved syntax highlighting.
sometimes the biggest announcements produce the smallest results.
English
Hidden cost of @PrajwalTomar_'s thread about Rork replacing Xcode: you still need to understand Swift. AI tools generate code, but someone needs to maintain it. The "no coding required" promise breaks down when the AI can't debug production issues.
x.com/PrajwalTomar_/…
Prajwal Tomar@PrajwalTomar_
You don’t realize how BIG this is. Rork can now generate fully native iOS apps using Swift, not React Native. It literally replaces Xcode. You can test instantly, install in one click, and ship faster than ever. 2026 is going to be the year of mobile apps.
English
gemini 3.1 pro: $2 per million tokens.
claud opus: $25 per million tokens.
12.5x price difference means gemini doesn't need to be 12x worse to win.
just needs to be good enough for your use case.
English
pascal levy-garboua: 6 micro-saas → $120k MRR → shut down 1, sold 2, kept 3.
portfolio approach beats betting everything on one product.
you get multiple shots and faster learning cycles.
English