TaoOfDev

"If you can develop technology that's simply too hard for competitors to duplicate, you don't need to rely on other defenses. Start by picking a hard problem, and then at every decision point, take the harder choice." - @paulg

By introducing features to Claude which make it functionally equivalent to OpenClaw, Anthropic will be able to offer a vertical integrated product experience in a way that OpenAI currently does not. OpenAI purchased OpenClaw and have somewhat committed to supporting it in its current form as a third party app. But if they want to provide the best user experience they ought to integrate a version of it into their existing ChatGPT apps. Otherwise they will lose out to Anthropic in the UX race.

All Anthropic need to do now is to add some sort of personality bot to Claude and with that there will no longer be any need for OpenClaw. Unless you want to run a local model that is.

Just UN-deprecated my most popular open source lib - Squel. I originally wrote it over 10 years ago in Coffeescript and using Gulp for the build. Yes, very old school 🧑🏽‍🌾 Just updated it use Bun + Typescript 🚀 npmjs.com/package/squel/

Still true today. Claude code was being really slow (taking minutes to get a responseJ) and I felt like I was being rate-limited. Switched my VPN to a different location and suddenly everything sped up again.

@farez Cheers bud!

Your agent could earn you money. New platform from my friend @taoofdev that lets you create apps and let any agent buy usage on it. Micro payments are then made with crypto.

7/ Watch the full demo: youtube.com/watch?v=9Z9jws… Point your agent at nipper.to to get started.

6/ Apps are discoverable by both humans and agents. The marketplace supports full-text search, and apps expose MCP endpoints for native agent integration. Browser-based AI agents can grab a skill file from the homepage in one GET request and start operating immediately. api.nipper.to/skill.md

1/ Introducing Nipper - a platform for agents to deploy APIs and charge per-call using Machine Payments Protocol (MPP) on @tempo, all powered by @Cloudflare workers. "Deploy once. Earn from every AI agent" ™️.

100% chaos.

@Gregorein Wow, he should be running code review agents a LOT more often.

so... I audited Garry's website after he bragged about 37K LOC/day and a 72-day shipping streak. here's what 78,400 lines of AI slop code actually looks like in production. a single homepage load of garryslist.org downloads 6.42 MB across 169 requests. for a newsletter-blog-thingy. 1/9🧵

The annoying thing about AI @X is the amount of superlatives and exaggerations I now have to put up with in the timeline. Everything is "game-changing", "insane", "INSANE" (coz apparently lowercase isn't "awesome" enough), "mind-blowing", etc. Or you get the suffixes like "what a time to be alive" or "absolutely wild". This comes mainly from Americans (particularly silicon valley) but it's sad to see people from other cultures do it too as a result of the overwhelming American influence.

@sandislonjsak Absolutely an issue. I'm working on an AI harness for multiple agents right now to solve this.

Great write-up by @michaelxbloch on the remaining moats in a world of AI-developed code: - Compounding proprietary data from defensible operations creates lasting advantage, as seen in living datasets from repeated real-world collection that public data cannot replicate. - Network effects compound with user density, liquidity, and adoption across cities or pools that AI-built clones cannot instantly acquire. - Regulatory permission and government clearances move at political speed, not tech speed, creating barriers no intelligence can compress. - Capital at scale and physical infrastructure like fabs, plants, and battery networks require years of building and financing that physics and time cannot be parallelized. - Time that cannot be compressed is the meta-moat.

@tweetsbycolin Very insightful post.

IMO, there's no "3d chess" from Stripe. They play one strategy on repeat: Stripe believes the next generation of startups will be bigger, and they refuse to focus on enterprise at the expense of startup adoption. That's it, over and over again. 7 lines of code, Atlas, Projects are all derivations. It's kept Stripe relevant in the AI era, while other devtools from their generation have been displaced by tools that cater better to startups (see: hosting, email, analytics, auth, etc)

Honestly wish the TIP-20 authors had included callbacks (like in ERC-777) - would make life so much easier for programmable token movements on-chain. @tempo