Fullstack developer @ Sybill.ai

Joined September 2015
132 Photos and videos
A top interview with a top guest, must watch We can call this interview "AI beyond the hype: what actually works"
New Weekly Dev's Brew is live. I got @badlogicgames properly grumpy about AI coding: why code is never free, why spec-driven dev is just waterfall again, and how he survives 50 AI PRs a day. Chapters 👇 0:00 Intro 2:20 Why Pi exists (and why Claude Code stopped fitting) 8:53 What success actually looks like for Pi 10:58 Building beyond the coding agent 13:13 Local AI is getting real 20:13 How Mario actually works 26:38 Discipline, atrophy, and juniors 29:08 Spec-driven dev is just hyper-waterfall 35:03 Code isn't actually free 37:56 Async agents and thinking time 44:42 Learning without the pain 49:13 The coming wave of sloppy software 52:11 GitHub under clanker load 53:31 Family, work, and balance 57:05 The Pi team and leadership 58:52 Refactoring Pi's core 1:03:32 Security, YOLO, and trust 1:06:27 Taming the PR flood 1:12:11 Token prices, and why Anthropic stopped courting devs
1
2
The Fable case should make you think how hard european countries are failing, we only think ways to screw each other to get some more money from EU central bank each year, we fumbled MERCOSUL, 28th regime, EU Army. In the most critical moment european countries failed to change.
18
The thing that i miss about Fable is how good it was at using the browser and organizing QA stuff, i have a self-improving harness that the agent improve a bit everytime I do browser QA, Fable was the first to organize and improve it properly, GPT 5.5 needs more guidance.
1
20
Fable had signicantly better organization, planning and jusdgement skills than GPT 5.5
10
The best thing about Fable 5 being taken down, is that GPT 5.5 is blazing fast
25
I achieved AGI minutes before Fable 5 got switched off.
42
Imagine now Open AI drops GPT 6
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Claude models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. Read our full statement: anthropic.com/news/fable-myt…
14
Mythos/Fable banned by the US government? My bet is they will revert tomorrow morning
12
Ray Thurn Void retweeted
WHAT?!
As a result of a US government directive, we are suspending access to Claude Fable 5 for all users. You can continue to use all other Claude models. Here’s what this means for you: Across Claude products, new sessions will run on your selected default model or Opus 4.8, and existing Fable 5 sessions will end with an error. On the Claude Platform, requests to Fable 5 will also return an error. Please update your integrations to other Claude models. We know this is a disruption to your workflows; we appreciate your patience and support.
3
1
114
19,810
multi threading in js?
13
Mythos released
It's out, Fable 5 model, most capable by a jump cursor.com/evals Just looking at the graph, Fable low is cheaper than Opus Max while being a bit better. Caveat is data retention for harm prevention (not training).
1
8
If you have a bash tool can perform safely basic aggregation arithmetic operations, like summing a column in a CSV No sandbox, no mythos, no "managed agents", just plain simple AI SDK
20
Imagine you have data in your DB, and you have an agent, and it sucks at reading it via tools and is so slow, and then you put bash in front and now it has superpowers, no sandbox, no VMs, no latency, no mounting S3 buckets with rmount, just plain old server-db rounds,
1
5
This is what is possible with just-bash @vercel, the world is not ready for it, a bash interepreter in javascript is really a superpowers your AI should have
4
War declaration to Remix 3
I’ve been experimenting with a new Inferno. It’s been 10 years since Inferno took off, and what better way than making it support hooks, TSRX and provide a level of performance that is incredibly consistent. It shifts the paradigm too, rather than having component re-rendering, we have control flow re-rendering. What if we didn’t create virtual DOM, but use something better. No signals, no blockdom, no signal-like API, just top-down rendering, but also something no one has considered before, or even explored. It might be an interesting experiment just purely to go against the common trend of being fine-grained. Maybe you don't don't need fine-grained rendering, when the top-down rendering is mostly just data, and not UI. Backed by an almost 1:1 API with React, minus RSC. It seems very promising. More to come soon.
9
$ Loop engineering is the shift from prompting AI agents one task at a time to designing systems that prompt, check, and steer them for you. Instead of being the operator in every turn, you build a recurring workflow: automations find work, worktrees isolate parallel agents...
1
2
35
skills preserve project context, connectors reach real tools, and sub-agents verify the output. The engineer’s role moves up a level: define the loop, set the memory, review quality, and decide what deserves trust. It’s not autopilot. It’s leverage with accountability.
1
2
Build the loop, but stay the engineer who understands what ships and why it matters.
1
The X bubble is always 3/6 months ahead of other platforms
4
Ray Thurn Void retweeted
Jun 8
i dont even write loops anymore my agents just read my mind while i phase shift into higher dimensions
24
26
366
26,071