Sergio R.

Builder.

+

[Sergio R. @TheRealSergio91]

https://t.co/uJ5F02Bq8l

[Sergio R. @TheRealSergio91]

engineering teams don’t have a tooling problem

they have a memory problem

today we’re relaunching Rivora on Product Hunt

open source. local first.

[Sergio R. @TheRealSergio91]

one thing i love about open source is how quickly papers become working implementations

turning research into something people can actually build on is one of the fastest ways the ecosystem moves forward

[Junu Park @junupark_]

happy to share what i've been working on lately - introducing tiny-speculators!

inspired by Speculators built by @RedHat_AI and @vllm_project team, i wanted to understand the end-to-end process of what it takes to train an EAGLE-3 draft model (for Qwen3-8B) from scratch.

link: https://t.co/cjCYyBupJB

👇checkout thread for more!

[Sergio R. @TheRealSergio91]

i love papers like this

everyone focuses on bigger models

meanwhile there’s a ton of innovation happening in the infrastructure underneath them

[Elliot Arledge @elliotarledge]

This is an absolute gemm in GPU collectives. Literally came out 3 days ago.

[Sergio R. @TheRealSergio91]

Every engineer should understand MVCC

It’s one of those foundational concepts that completely changes how you think about Postgres

Another fantastic write up from the PlanetScale team

[PlanetScale @PlanetScale]

VACUUM is notoriously difficult to get right for busy Postgres databases.

The reason it's needed boils down to MVCC and how Postgres implements it.

Our latest blog unpacks how it all works.

[Sergio R. @TheRealSergio91]

this is a real move from meta

agentic performance + long context + aggressive pricing is going to matter

[Mark Zuckerberg @finkd]

(1) Today we're releasing Muse Spark 1.1 -- a strong agentic and coding model at a very low price. It's available through our new Meta Model API and in Meta AI.

[Sergio R. @TheRealSergio91]

really interesting shift

once evals, rules, and alerts are provisioned like infrastructure, agent reliability starts looking a lot more real

[Harrison Chase @hwchase17]

Setup evals with… terraform??

[Sergio R. @TheRealSergio91]

directionally agree

feels like agent execution is moving from full linux environments to the smallest runtime that still gets the job done

[Nathan Flurry 🔩 @NathanFlurry]

🔮 Agents won't run in Linux sandboxes in 3 months

Worth watching:

- @LangChain's experiments
- @rivet_dev's agentOS for WASM-based sandboxing
- @CloudflareDev's Project Think using V8 isolates
- @vercel_dev's just-bash for Bash-in-JS

Links below 🧵

[Sergio R. @TheRealSergio91]

this is interesting

feels like the real shift is not “no chat”

it’s moving from chat as the primary interface to chat as one view into a bigger system

[Chris Tate @ctatedev]

The future of agentic coding and review is chatless

Chat becomes a projection layer: logs, traces and context for o11y and audits

The primary interface becomes dashboards that expose goals, progress, blockers, confidence and outcomes

[Sergio R. @TheRealSergio91]

really important release

making vllm + lmcache hackable on a macbook with no gpu lowers the barrier a lot

[LMCache Lab @lmcache]

𝐍𝐨 𝐆𝐏𝐔? 𝐍𝐨 𝐩𝐫𝐨𝐛𝐥𝐞𝐦.

We just published a starter guide for developing vLLM + LMCache on a MacBook.

LMCache's multi-platform design decouples the GPU from most core data paths, so a single laptop is enough to clone, build, run unit tests, and verify a real cache hit on CPU. The guide walks through the environment setup in ~10 minutes and points to four concrete areas where you can start contributing.

If not having a GPU was your only blocker, it is not anymore.

Read the guide and join us in building the KV cache layer for faster LLM inference: https://t.co/l22JWaH22F

#LMCache #vLLM #LLM #AIInfrastructure #OpenSource

[Sergio R. @TheRealSergio91]

Root is live on Product Hunt today 🚀

would love your feedback!

https://t.co/L66kNurn5e

[Sergio R. @TheRealSergio91]

package managers install software

agents will install entire systems

trust is about to matter a lot more

that's why i built Root

https://t.co/0Fd5eaAOdq

[Sergio R. @TheRealSergio91]

I've been cooking 🔥

[Sergio R. @TheRealSergio91]

Built @Arbora_AI on Cloudflare

AI that understands the physical world, starting with contractors

Type an address → get property analysis, opportunity scoring, AI-generated concepts, and homeowner-ready project packets.

Powered by Workers, D1, R2, KV, Queues, Workers AI, and Durable Workflows.

[Sergio R. @TheRealSergio91]

@boardyai Building @Arbora_AI

Contractors are spending thousands chasing leads that can’t even fit a pool.

We’re fixing that.

[Stu Holden @stuholden]

HAPPY WORLD CUP DAY. LFG!!!!!!

[Sergio R. @TheRealSergio91]

@JunaidAckroyd Building https://t.co/zdiK0rQjGc

[Sergio R. @TheRealSergio91]

@16vchq https://t.co/zdiK0rQjGc

[Sergio R. @TheRealSergio91]

Here's Root today

Still early, but seeing it go from podcast idea → working software has been incredibly fun

v0.1.9

[Sergio R. @TheRealSergio91]

thank you for the shout out on the pod @mattzcarey @wilhelmklopp

Shipped multiple releases and Root is now at v0.1.9

Funny enough, building Root also changed how I think about AI products entirely

[Sergio R. @TheRealSergio91]

Root started as a simple idea

what if package managers managed your entire machine instead of just packages?

install tools
sync environments
reproduce machines
all from a single lockfile
https://t.co/CVIlDdpUl9