pepmach

Building Kiro Crew @AWS

[pepmach @pepsi_machine]

Kiro Crew is live & fully open source TODAY!!!

https://t.co/ArCzTZ1MOJ
https://t.co/fU8FLOA5hS

1) Memory that compounds — corrections become lessons, repeated workflows become skills. Not a context file you hand-curate; it writes itself, in Markdown you can inspect, edit, or delete.

2) Loops, not prompts — crons, webhooks, heartbeat monitors, parallel subagents. CI fails at 3am, you review a fix at 9.

3) Multiple channel support — Slack, Discord, Telegram, WeChat. Each thread is its own session; steer a running task from your phone.

4) Apps — for the work chat can't carry: wrap agents, skills, and schedules in your own UI. SDK in TypeScript or Python.

5) Any Kiro-supported model — Claude, GPT, GLM, Qwen, pick per task. Your skills from other harnesses drop in as-is.

6) Security you can audit — sandboxing, credential redaction, signed audit logs. It's all in the repo.

Battle-tested by 40k+ builders inside Amazon before today's public release.

[Jeff Dean @JeffDean]

Announcing Discovery Loop!

I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the world’s most used products, infrastructure and AI models, and we’re excited to turn our attention to this ambitious endeavor.

♾

Learn more at: https://t.co/Rv3LMdLluK

[pepmach @pepsi_machine]

@dylan522p what’s the secret part about this

[pepmach @pepsi_machine]

AI can do better science discovery/automated research better than Nobel prize laureates

i place my bets on pharmaceuticals

[Charlie Greenman @chargrnmn]

@karpathy I wonder what the equivalent of an "ahead of time prediction" like this today is?

[pepmach @pepsi_machine]

@saranormous for me it’s terra on xhigh. feels like a good balance of speed and intelligence

[pepmach @pepsi_machine]

@FunnyAstro @Viol2t_OW YOU LITERALLY DIED TO WBG BASTION ULT ON JUNO LMAOOOOO

[pepmach @pepsi_machine]

LOOOOL this is the most SF thing ever

[Saai Arora @SaaiArora]

Drake concerts in sf go crazy

[pepmach @pepsi_machine]

@zephyr_z9 i don’t think we can help ourselves

[Kimi.ai @Kimi_Moonshot]

Releasing the model weights and technical report of Kimi K3.

Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.

New model architecture: 2.5x the intelligence per unit of compute, not just more params.

Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.

Model weights: https://t.co/7m7eEg6Y0B
Tech report: https://t.co/yeu6cjpMCT
Tech blog: https://t.co/YTfiMSNM1f

[pepmach @pepsi_machine]

@jukan05 Are getting deployed to Moonshot AI or wut

[pepmach @pepsi_machine]

peak 🏔️

[Sonya Huang 🐥 @sonyatweetybird]

The first Vera Rubin clusters are here!

Yesterday, @IneffableLabs took delivery of their Vera Rubin NVL72 cluster from @googlecloud @nvidia

The AI frontier jumps forward by yet another generation of hardware. Acceleration continues.

[Andrej Karpathy @karpathy]

One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 minutes, total mess, anything goes, full stream of consciousness. Sometimes I declare it up top, something like "switching to speech recognition sorry for any typos...". Sometimes I turn it into a small interview of a few turns. But I find that the LLMs are somehow very good at reconstructing long incoherent rambles and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with. The result is that you improve the mind meld and have to correct things less from that point on.

[pepmach @pepsi_machine]

What flags? All I'm seeing is token [:es:] got translated into token [:us:]

[pepmach @pepsi_machine]

they said flying cars tho but instead i’m getting better dating advice from gpt-5.6-sol than from therapy

[Demis Hassabis @demishassabis]

https://t.co/PTeDiv1b6L

[John Carmack @ID_AA_Carmack]

Memory cost and capacity are significant issues for AI accelerators.

Unlike game rendering, model inference can have a deterministic memory access pattern. You don’t need “random access memory” at all for model weights, and you could tolerate cold-start latencies in the multiple milliseconds, as long as continuous reads were delivered at the necessary bandwidth.

NAND flash is over 100 times cheaper per GB than HBM, so there should be opportunity there, even after giving a flash controller a 1024 bit interface with HBM bandwidth.

You could make a specialized pin protocol that just supported pipelined transfer of full 16KB+ pages from the flash to program-managed accelerator scratchpad memory and improve per-pin performance over HBM, but it might be more convenient to make it still look like a true random access memory with very fragile performance characteristics, where anything but sequential reads falls off a 1000x+ performance cliff.

That has the advantage of automatically using existing cache hierarchies, and providing a natural path to update the flash memory with new model weights. With the stream-to-scratch interface, code has to be completely rewritten before it works at all, while the ram-emulation interface will start off just extremely slow, and you can incrementally sort out the changes for full performance.

There may be cases where there isn’t enough scratchpad SRAM to hold the weights for a layer, which might force you to deploy the old optical drive optimization technique of duplicating data in multiple places on a sequential read to avoid seeking, but there would be capacity to burn.

It might be possible to do something like cuda graph capture to record a memory access trace and have everything magically remapped to a linear sequence, but deploying programmer / agent elbow grease to manage transfers and access in a scratch ram ring buffer would be lower risk.

A split memory system consisting of some channels of flash and some channels of HBM will probably be suboptimal compared to a uniform memory, but it could be much cheaper, and allow much larger models to be run.

I think th case is strong for inference, but you have to stretch more for training. You can still linearize all the weight memory accesses, both reads and writes, but flash memory would quickly wear out from the writes, even if they were all perfectly page aligned. Replacing low-latency HBM with massively parallel cheap(er) DRAM at high latency might still be a worthwhile cost savings.

[pepmach @pepsi_machine]

@jukan05 as long as they’re open weights and i can host them on a tinybox then i don’t have problem with them.
i don’t understand why you put quotes around open-source.

[Boris Cherny @bcherny]

As engineering, product, design, DS, etc. melt into a new kind of role, I was reflecting on what roles might look like in the future. For example, when I look at the Claude Code team I see what I think is five archetypes:

1. Prototyper: comes up with brand new ideas; churns out many ideas, most of which don't ship
2. Builder: quickly turns a prototype/idea into production-grade product/infra
3. Sweeper: cleans up the UI, simplifies the code and system, unships, optimizes performance
4. Grower: takes a product that has been built and iterates on it to improve Product-Market Fit
5. Maintainer: owns a mature system to make it secure, reliable, fast, and efficient as it scales

Many people span across 2 roles, and sometimes 3 roles. I also notice that these roles are not really tied to job function -- eg. across Anthropic, some designers match category 1, some 2, some 3; same for engineers, PM, DS.

A healthy team needs a mix of these, depending on the product:

- A product that is new and pre-PMF needs people that are strong at 1+2+3
- A product that is growing and has found PMF needs 2+3+4 and some 5
- A product that has strong PMF needs 3+4+5 and some 2

Maybe product roles of the future will look more like this, and less like the domain-specific roles of today?