Jordan Bryan

@CausalMarkets @arpitrage - Codex is already the better harness at the moment - OpenCode already exists as an OSS alternative to CC. It's fine. - Considerable amounts of this code will be rewritten in the coming year anyway. These products are being improved at an astounding rate.

@jobryan205 @arpitrage All there learnings are public

@ZachAbramowitz theredline.versionstory.com/p/why-cant-43b…

Why do law firms choose Harvey and Legora over Claude? Right answers only.

@david_perell Is it any more hollow and corporate than Starbucks or any other chain, for that matter?

Blank Street is the coffee shop you'd make if you asked ChatGPT to create a brand for you. The Helvetica font, matte green color, and even the word 'Blank' are designed to appeal to everybody, everywhere and as hollow as the brand may be, you have to admire how well-packaged it is as an investment vehicle. Every time I take a sip of Blank Street latte, I think about how its success says more about our modern world than just about any other up-and-coming brand. Every critique you make of Blank Street is a critique you can pass onto our society writ-large. You can't help but appreciate the brand once you think of it as modern art.

@adrianasobota_ @gavinpurcell I disagree on the UX front. Codex desktop app is the best coding agent UX right now IMO. Claude Code was better UX than Cursor immediately upon release a year ago. OAI and Anthropic don’t have a Replit competitor yet, but Google AI studio is great!

The bull case is that model providers are terrible at UX and distribution. OpenAI has tried to ship coding tools for two years and Cursor still grows faster. Anthropic ships Claude artifacts and Replit still has more paying users building real apps. The model layer commoditizes but the workflow layer compounds. Every integration, template, and deployment pipeline these companies build is a switching cost that raw model access can't replicate.

still having a hard time understanding how companies like replit, bolt, lovable, cursor keep their valuations i understand they’re documenting huge arr etc but how do they not get eaten by anthropic and openai? what is the bull case there? user lock-in? features?

@mr_james_c That observation is perfectly aligned with his thesis in the essay

I started my career working on tobacco brands. There's virtually zero product differentiation. Yet smokers would define their personality based on the brand they smoked. So I can tell you this article is 100% wrong.

@amoljain_ @toddsaunders How does Replit beat Google AI studio?

@toddsaunders Find someone non-technical in your life, give them Claude Code and ask them to ship a production grade app. You’ll quickly find out why Replit exists.

No shame on Replit, built a bigger business than I've ever built. But at the rate Claude Code /Cowork is moving, why does Replit needs to exist? What's the bull case for the future of Replit and similar tools if all the model companies are on the same path.

@jarredsumner Why did you raise a Series A if you only spent 20% of your seed?

i wouldn’t suggest fundraising < $4m. we did $7m for bun’s seed round and it was probably too much - I don’t remember exact number but I think we spent like 20% of that by the series A.

now that I’m no longer doing a startup and won’t for many years, some early stage startup fundraising advice: don’t spend any time at all with investors until you’re ready. tell them you’re too busy. do not meet with them. yes especially if an associate emails you 5 times

@tszzl @kimmonismus Is confusing naming a deliberate strategy on OpenAI's part?

@kimmonismus it’s extremely confusing even for me

Seriously, I dont get it. - Today, GPT-5.3 instant is being released . - The blog post states at the very bottom that 5.3 Thinking and Pro will also be released very soon. - An hour later, the official OpenAI X account tweeted that GPT-5.4 will be released very soon. ??? So in a few days we get: GPT-5.3 thinking + Pro + GPT-5.4 (???) instant/thinking/pro?

@Yashraj__ @zackbshapiro "Legal AI" companies are using model providers under the hood with the same zero-day retention policies

@zackbshapiro > Anthropic offers a zero-data-retention API option and business data processing agreements, so that none of your client data is used to train models, and inputs are not stored beyond the session. "trust me bro" maxxing? 🥶

@sbcatania @NotionHQ I guess I'm curious, on a conceptual level, what the advantage of running a worker in Notion is compared to Claude connecting to the same systems via MCP.

@jobryan205 @NotionHQ I'm not sure how the spotify MCP works but you might be able to! this is just the beginning of what workers can do. though you can already do much more than just build an integration with workers — since they can run any custom code, you can do much more

today, on the DL, @NotionHQ released WORKERS. workers are a generic code execution environment that let you build interfaces within Notion the first interface we're exposing is the ability to make your own custom agent tools you can now make a tool for a custom agent that executes arbitrary code... and if you can execute arbitrary code, you can do pretty much anything!

@ramez @mattyglesias Can you elaborate? What's it saying?

@mattyglesias I'm bullish on AI and deeply skeptical of the METR graph. It isn't saying what most people think it's saying.

@MattZeitlin I published an essay today in response these types of criticisms! I figured the best way to show people how pointless that form of criticism is is to show them what Claude Code can do concretely. theredline.versionstory.com/p/dont-trust-p…

whenever people downplay AI capabilities by saying "actually it's just a next token predictor" all i can think of is

Full essay here: theredline.versionstory.com/p/dont-trust-p…

But if you've driven down a highway at 70mph, you know that it's so irrelevant, that it's frankly not worth discussing.

While AI critics pontificate about whether AI is *truly* intelligent, here are three concrete things I've done with Claude Code that have transformed the way I work:

The outputs of these exercises pertain to software startups, but the problems are general. Analyze data to create a report. Find precedent answers for a compliance form. Build a tool for a workflow the market doesn't serve. Many roles deal with problems like these.