Alan Oliver

Building in public with Claude Code & automating with @OpenClaw.

[Alan Oliver @AlanOliverDev]

I just moved to the completely new opened Lunga Coliving, Siargao where I’ll be staying for the next few months! I just spent 3 hours creating a @openclaw bot to collect the newest residents feedback for my gf who is working as community manager

[Alan Oliver @AlanOliverDev]

Suddenly I’m starting to think about moving back over to Claude.
Bloated over abstracted functional and much too defensive code with codex. There’s room for both, especially when building complex architecture

[Claude @claudeai]

New in Claude Code: Artifacts.

Interactive pages built from your session, like a PR walkthrough or a living project dashboard, shared with your team at a private link.

Available in beta on Team and Enterprise plans.

[Elon Musk @elonmusk]

Where will AI be in 1, 2 or 3 years?

[Alan Oliver @AlanOliverDev]

Chinese gamifying US frontier company models

[aditya @adxtyahq]

Chinese students are buying GPT-5.4/5.5 and Claude API access from Xianyu/Taobao proxy sellers for almost 96-97% cheaper

People are apparently burning 100M+ tokens a day for like $1 and vibecoding nonstop.

[aditya @adxtyahq]

Chinese students are buying GPT-5.4/5.5 and Claude API access from Xianyu/Taobao proxy sellers for almost 96-97% cheaper

People are apparently burning 100M+ tokens a day for like $1 and vibecoding nonstop.

[Alan Oliver @AlanOliverDev]

Software scaling meets hard reality in a AI GPU hungry era

[Hedgie @HedgieMarkets]

🦔A data center in Fayetteville, Georgia, drained approximately 30 million gallons of water through two industrial-scale hookups that the local utility did not know existed. One connection had been installed without the utility's knowledge, and the other was not linked to any account and therefore was not being billed. The discovery only came after residents complained about low water pressure.
The campus is still under construction with completion projected three to five years out. A separate incident in Tucson last week saw Project Blue's contractor caught trucking municipal water out of a city that had explicitly voted against the project, with Tucson revoking the temporary meter and demanding two acre-feet of water credits to make the city whole.

My Take
Two unrelated data center water incidents in two weeks across two different states is a pattern, not a coincidence. The Georgia facility was running off an unmetered industrial hookup nobody at the utility had on file, which means either a contractor installed it without authorization or the utility lost track of a connection serving a major customer, and neither of those explanations should make anyone comfortable. The construction phase alone consumed 30 million gallons before operations even began, which gives you a sense of the water demand profile these facilities have once they go live.

The bigger issue is that hyperscale data centers are being permitted under regulatory frameworks built for industrial users a fraction of their size, and the utilities responsible for tracking water use are not staffed for facilities this scale. A 30 million gallon discrepancy slipping through billing is not a clerical error, it is a sign that the infrastructure for monitoring these projects is being outpaced by the speed at which they are being built. Tucson caught their problem because a citizen made a phone call to a council staffer, and Fayetteville caught theirs because neighbors noticed their taps had lost pressure. Neither of those is a functioning compliance system, and the next community in this situation will probably not catch it at all.

Hedgie🤗

[Alan Oliver @AlanOliverDev]

Built using Hermes agent and bird cli!

[Alan Oliver @AlanOliverDev]

Today I shipped lingonudge - X topics of your interest can be summarised and turned into a listening class in whatever language.

https://t.co/C1fbdRUgf6

[Alan Oliver @AlanOliverDev]

I use it myself for Mandarin intermediate listening.

Give it a go completely free! I’d love some feedback!

[Rik · midcurved 🟨 @rikventure]

if i had to recommend 1 podcast to listen this weekend it would be this one

original takes of real builders on everything what's happening in AI land

no AI slop, just good banter & the latest news

https://t.co/ivtXWLcYka

[Alan Oliver @AlanOliverDev]

So impressive! Too many tools to try and not enough time

[Marko Slavnic @Markoslavnic]

The quality of animation you can create on your own is truly amazing. We really are just limited by our imaginations at this point. Go tell your story!

Made in @runwayml in a few hours and a handful of gens.

[Alan Oliver @AlanOliverDev]

@5minaid 嗯 我同意。比如我Hermes agent 不允许exec 工具
允许我自己做非常有限的mcp tool权限给的model

你说的例子很有应景

[Alan Oliver @AlanOliverDev]

@rikventure I’m gonna start making specs html files not markdown, way more easy on the eye. Especially for architectural diagrams

[Alan Oliver @AlanOliverDev]

Suddenly datacenters in space proposed by Elon doesn’t sound like such a crazy idea. Natural cooling in orbit

[Brian Armstrong @brian_armstrong]

We experienced an outage at Coinbase last night, which is never acceptable. The root cause was a room overheating in an AWS datacenter when multiple chillers failed. We design our services to be redundant to downtime in any one AWS Availability Zone (AZ), and most of our systems worked this way last night, but not all.

Our centralized exchange did not. Exchanges have unique architectures that optimize for latency and co-location of clients. It is possible to make exchanges resistant to AZ failures, but this can introduce latency delays that are not desirable along with breaking customer co-location. Given this incident, we'll revisit these tradeoffs to ensure we're giving you the best possible venue to trade. At a minimum, the duration of an outage should be able to be reduced considerably when an AZ move is needed.

Thank you to the AWS and Coinbase teams for working through the night to mitigate the issue. We’ll share the detailed technical summary once it's ready.

[Rik · midcurved 🟨 @rikventure]

I asked Codex to create a meme about the SpaceXAI x Anthropic news and it did not dissapoint

[will brown @willccbb]

hey mind reviewing these real quick

[Alan Oliver @AlanOliverDev]

Made the switch

[anita · vellum.ai 👾🦾 @anitakirkovska]

it's insane how most people on my feed went from claude to codex in just 2 days. this industry is crazy

[Alan Oliver @AlanOliverDev]

@yrzhe_top Nice idea. I’d like to see an execution trace and tool calling where it’s easy to see what decisions agents are making in an intuitive way

[Alan Oliver @AlanOliverDev]

@wasadigi @guiassisbrasil How have you tried to do the handoff so far? Sounds like you’d have to definite frames and annotations for each shot.
Or is exporting the artifact and that’s enough context for Claude code? I’d imagine more context is needed to definite what a “match” means

[Alan Oliver @AlanOliverDev]

Speed at a baseline is determined by latency of the model provider you’re using. And then depends on the complexity of request.
Tool calling like reading files or a web fetch will take time. Oversized repo and context bloat can increase that further. So the answer is a mix of those

[Alan Oliver @AlanOliverDev]

@web3nomad @steipete Looks like the GitHub action runner agent has context per run of a readme summary and items / closed folder. So it doesn’t go in cold, but those summaries will have some information from previous runs