Rendex retweeted
Claude Code creator:
"The model just keeps getting better. Everything around the model is also increasing exponentially."
One setting feels like cheating.
Most people have never opened it.
7 more like this are in the article below. All free. All already in Claude. All sitting there the whole time.
Bookmark this before you close the tab.
Rendex retweeted
Andrej Karpathy, on building llm.c:
"The LLM could act as a kind of compiler for any custom application you're interested in. They're going to output the binary you can compile and run."
He said this two years ago, it sounded theoretical.
Now Claude Code ships Artifacts, a session turns directly into an interactive page, a PR walkthrough, a living project dashboard, shared with your team at a private link.
No separate build step, no deploy pipeline, the session itself becomes the product.
Beta on team and enterprise plans.
The thing Karpathy predicted is just shipping now, one feature at a time.
Rendex retweeted
Boris Cherny (Claude Code creator, Anthropic):
"Engineering has never been this fun. I feel like I suddenly have superpowers. I have a jetpack."
He's not talking about autocomplete.
He's talking about understanding the stack underneath.
Tokens, context windows, agentic loops, stop conditions, evals, context engineering.
The 6 concepts that separate people who use AI from people who build with it are in the article below.
Bookmark it.
Rendex retweeted
Boris Cherny (Claude Code creator, Anthropic):
"I don't prompt Claude anymore. My job is to write loops."
One day he wrote a diary of everything he did. Highlighted recurring tasks with different colors. Each color became a loop.
That exercise is worth more than any $500 course on agents.
Watch the workshop below ↓
Rendex retweeted
Diane Penn, Anthropic (shipped 21 versions of Claude):
"Instead of asking Claude to write a project update, you can now ask Fable to make sure the project stays on track for the whole week."
That is not a prompt, that is a loop.
Fable 5 can run for days on a single goal without losing the thread. It dispatches sub-agents, keeps them on track, and remembers your specs across millions of tokens.
The shift Anthropic is describing is exactly what the article above breaks down: stop typing tasks. Start building systems that run them.
One prompt gets you one answer. A loop gets you an agent that works while you sleep.
Anthropic engineer:
"you're not supposed to prompt Claude. you're supposed to build a system that prompts itself [loops]."
this is one of the best workflows I've seen in a long time
in this video he breaks down exactly how most people are building loops wrong:
- the memory file you never set up, so every loop starts from zero
- the sub-agents that 95% of builders have never split apart
- the stop condition setup that keeps loops from running forever and billing you in your sleep
- why writing one prompt a day is the slowest way to use Claude
if you've been using Claude for more than a month and still typing every task by hand, you've been running one prompt when you could be running a system of loops
instead of another prompt tonight, watch this
make sure to bookmark it before it gets buried
full guide in the article below
Rendex retweeted
Boris Cherny (Claude Code creator, Anthropic):
"At any moment in time I have a few agents, or sometimes thousands of agents running, doing my work. I can think of ideas, talk to customers. They do the part I don't enjoy."
That is not a demo. That is production.
One agent hits a ceiling. A coordinated system of specialists breaks through it.
94.3% first-pass success rate on 1,847 daily runs.
That is the number worth building toward.
Full pipeline architecture below ↓
Rendex retweeted
Jess and Michael (Anthropic, Claude Managed Agents) at Code with Claude:
"The bottleneck is increasingly the infrastructure, not the intelligence."
Two years ago you prompted Claude to write a component. Last year you debugged files for an hour. Now agents run overnight across agent teams and action your entire backlog before you wake up.
The infrastructure gap is what Managed Agents closes.
9 out of 10 developers building agents today are rolling their own harness, managing their own context, and building their own observability from scratch.
Anthropic did the platform work so you don't have to.
Read the full agent build guide below.
Worth more than a week of custom infrastructure.
Rendex retweeted
Fable 5 just shipped dynamic workflows and the framing isn't "run more agents," it's "load the board and walk away."
Drop your project tickets into GitHub. One command. The orchestrator spins up builders, QA agents, and reviewers in a loop until every ticket is done.
One developer ran it on a full lead magnet app: 11 tickets, all processed autonomously, PRs merged, board empty.
Six patterns under the hood that handle any type of work:
classify, fan-out, adversarial review, generate-filter-tournament, loop until done.
The one-agent-one-task era is closing.
The agentic OS just opened ↓
Rendex retweeted
Dario Amodei (CEO, Anthropic) at Davos:
"The team that leads Claude Code says he hasn't written any code in the last two months. It's all Claude. He's edited it, he's looked at it, but it's all been written by Claude."
That is what autonomous agents look like when you actually build the loop correctly.
Fable 5 + self-correction + verification loops.
9 out of 10 developers still babysit every step.
The CEO of Anthropic is describing a world where the person who builds the product doesn't write code anymore and most people building with Claude haven't even set up a basic verification loop yet.
Read the 5-step build below.
Worth more than a month of manual checkpoints.
Rendex retweeted
Diane (Anthropic, shipped 21 versions of Claude):
"Instead of asking Claude to write a project update, you can now ask Fable to make sure the project stays on track for the whole week."
In 10 minutes on stage in Tokyo she reveals what Fable 5 actually changes for developers building with agents.
Fable 5 + multi-agent pipelines + long-horizon autonomy.
9 out of 10 developers are still building single-agent wrappers around Claude.
The person who ships Claude is already talking about agents that run for days, dispatch sub-agents, and validate their own work.
Read the full pipeline breakdown below.
Worth more than any multi-agent course you'll find online.
Rendex retweeted
Thark, Anthropic Claude Code team engineer:
"I often find Fable produces better code than me. My job is increasingly about direction and setup, not supervision."
That's not a marketing line. That's the actual change in workflow.
Here's the 3-shift framework he uses with Fable 5:
• context not constraints. instead of "keep it simple," say "this feature might be deleted in a month, don't build anything painful to throw away"
• goals plus verification. set a goal, add a workflow to verify each part, let Fable run
• raise the ambition bar. if you assumed Claude couldn't do it, try it now. Thark is editing video with Fable
The model routing config that handles this automatically is in the article below ↓
Rendex retweeted
Ivan Nardini, Google Cloud developer advocate, on stage with a room full of engineers:
"I just drew the app idea on paper. Claude rendered the wireframe, built the API, ran the security review, and deployed it - in one session."
24 minutes where Ivan ships what normally takes weeks of PM → design → engineering → security handoffs
One live demo worth more than a sprint planning meeting
Ivan Nardini, Google Cloud developer advocate, turned a paper sketch into a live production app in one session.
No back and forth. No design cycles. No waiting on engineers.
Here's how the full build went:
• PM draws the app idea on paper → Claude renders a working wireframe from the photo
• UX: plan mode full component list proposed before any code gets written
• engineer: subagents build API, pipeline, and dashboard in parallel
• security: auto review, auto fix, auto deploy to Cloud Run
The old way takes weeks of handoffs between PM, designer, engineer, and security while the teams who get this compress the entire cycle into one afternoon
Watch the full demo, then read the article below ↓
Rendex retweeted
Boris didn't change his workflow. He changed his model, and his output went from 10% AI to 100%.
This is the clearest signal yet that we're not in a "AI helps you code faster" era anymore.
We're in "the loop runs, you design the loop" era.
That's not a productivity gain. That's a role shift.
The engineers winning right now aren't the ones prompting Claude the best. They're the ones writing systems that prompt Claude for them.
And that distinction between using AI and orchestrating AI is going to define who builds what in the next 12 months.
Worth 30 minutes of your time.
Boris Cherny, Claude Code creator, revealed what Anthropic actually looks like inside, and it changes how you should think about your team.
In this 30 min talk, Boris revealed the actual shift happening in how engineers work with AI.
Here's what he covers:
• Claude Code went from writing 10% of his code to 100%, the only thing that changed was the model
• ramp-up time at Anthropic dropped from weeks to 2 days, new hires just ask Claude
• the next abstraction: loops prompt Claude, not you, your job is to write the loops
• engineer vs PM vs designer, all roles merging into one builder by end of year
Most people are still prompting Claude manually every time - while the people who get this are writing loops that prompt Claude for them
Watch the full talk, then read the article below
Rendex retweeted
Boris Cherny, Claude Code creator, revealed what Anthropic actually looks like inside, and it changes how you should think about your team.
In this 30 min talk, Boris revealed the actual shift happening in how engineers work with AI.
Here's what he covers:
• Claude Code went from writing 10% of his code to 100%, the only thing that changed was the model
• ramp-up time at Anthropic dropped from weeks to 2 days, new hires just ask Claude
• the next abstraction: loops prompt Claude, not you, your job is to write the loops
• engineer vs PM vs designer, all roles merging into one builder by end of year
Most people are still prompting Claude manually every time - while the people who get this are writing loops that prompt Claude for them
Watch the full talk, then read the article below
