Jan

"I" between the User and Code

[Jan @uicoded]

@NoahEpstein_ DESKTOP

[Jan @uicoded]

@NoahEpstein_ How does: “your sensitive files never hit the cloud”?
Yes it can move, rename files without reading them, but that’s very limited use case. What else am I missing?

[Daniel Berk @danielcberk]

Very good beehiiv websites

https://t.co/eHHno9YVj7

[Jan @uicoded]

@darshal_ Yt

[Jan @uicoded]

@basicprompts Two shorter shorter sessions on different days beats same time on single day

[Govardhana Miriyala Kannaiah @govardhana_mk]

DevOps official language has to be 'YAML'

- Helm uses YAML
- GitHub uses YAML
- Ansible uses YAML
- Argo CD uses YAML
- Kubernetes uses YAML
- Azure DevOps uses YAML
- Docker Compose uses YAML
and more ...

All the sophisticated toolsets you aspire to learn and adopt run on YAML.

Get that right first.

55K+ read my DevOps and Cloud newsletter: https://t.co/WBucLdwLhJ

What do we cover:
DevOps, Cloud, Kubernetes, IaC, GitOps, MLOps

🔁 Consider a repost if this is helpful.

[Avi Chawla @_avichawla]

Researchers from Meta built a new RAG approach that:

- outperforms LLaMA on 16 RAG benchmarks.
- has 30.85x faster time-to-first-token.
- handles 16x larger context windows.
- and it utilizes 2-4x fewer tokens.

Here's the core problem with a typical RAG setup that Meta solves:

Most of what we retrieve in RAG setups never actually helps the LLM.

In classic RAG, when a query arrives:

- You encode it into a vector.
- Fetch similar chunks from vector DB.
- Dump the retrieved context into the LLM.

It typically works, but at a huge cost:

- Most chunks contain irrelevant text.
- The LLM has to process far more tokens.
- You pay for compute, latency, and context.

That’s the exact problem Meta AI’s new method REFRAG solves.

It fundamentally rethinks retrieval and the diagram below explains how it works.

Essentially, instead of feeding the LLM every chunk and every token, REFRAG compresses and filters context at a vector level:

- Chunk compression: Each chunk is encoded into a single compressed embedding, rather than hundreds of token embeddings.
- Relevance policy: A lightweight RL-trained policy evaluates the compressed embeddings and keeps only the most relevant chunks.
- Selective expansion: Only the chunks chosen by the RL policy are expanded back into their full embeddings and passed to the LLM.

This way, the model processes just what matters and ignores the rest.

Here's the step-by-step walkthrough:

- Step 1-2) Encode the docs and store them in a vector database.
- Step 3-5) Encode the full user query and find relevant chunks. Also, compute the token-level embeddings for both the query (step 7) and matching chunks.
- Step 6) Use a relevance policy (trained via RL) to select chunks to keep.
- Step 8) Concatenate the token-level representations of the input query with the token-level embedding of selected chunks and a compressed single-vector representation of the rejected chunks.
- Step 9-10) Send all that to the LLM.

The RL step makes REFRAG a more relevance-aware RAG pipeline.

Based on the research paper, this approach:

- has 30.85x faster time-to-first-token (3.75x better than previous SOTA)
- provides 16x larger context windows
- outperforms LLaMA on 16 RAG benchmarks while using 2–4x fewer decoder tokens.
- leads to no accuracy loss across RAG, summarization, and multi-turn conversation tasks

That means you can process 16x more context at 30x the speed, with the same accuracy.

The code has not been released yet by Meta. They intend to do that soon.

[Jan @uicoded]

https://t.co/oCxDslbOWN
Help for those devs not reading the @tailwindcss IntelliSense plugin docs.

[Jan @uicoded]

"VSCode CSS `unknownAtRule` warning fix"
for those not reading the docs 😆
@TailwindCSSDev @tailwindcss @uicoded #DEVCommunity https://t.co/oCxDslbOWN

[Jan @uicoded]

The Gang of Four (GOF) Part II → Agentic Design Patterns @antoniogulli

[ℏεsam @Hesamation]

Calculating GPU memory for serving LLMs

[Jan @uicoded]

https://t.co/RxJS4ByfJo

[Jan @uicoded]

@IgorMinar @Cloudflare Great post, worth revisiting.
It has been 3YRS - excited for follow-up.
+1 for the experiences viewpoint.

[Jan @uicoded]

@johnlindquist well deserved feedback.

[Joe Gebbia @jgebbia]

My directive is to update today’s government services to be as satisfying to use as the Apple Store: beautifully designed, great user experience, run on modern software. An experience that projects a level of excellence for our nation, and makes life less complicated for everyday Americans.

I’d like to thank President Trump and his administration for creating this vision, and supporting America by Design (@americabydesign). I will do my best to make the U.S. the most beautiful, and useable, country in the digital world. 

If you’re interested in joining me at the newly established National Design Studio (@NDStudio) on this ambitious mandate, please reach out with a link to your work. Thank you for your attention to the details.

[Joe Gebbia @jgebbia]

I’m honored to be appointed by President Trump as the first Chief Design Officer of the United States of America.  (🧵)

https://t.co/PPq5nb9ok1

[colinhacks @colinhacks]

it should be noted: everything that supports Standard Schema works out of the gate, including:

- tRPC
- TanStack Form & Router
- OpenAuth
- t3-env
- Orval
- FastMCP
- Content Collections
- cachified
- up-fetch
- renoun

and many others listed here: https://t.co/jrySHjWimF

[Wes Bos @wesbos]

ITS HAPPENING

Firefox 127 shipped today making it the first browser to support Temporal - the new API for working with Dates, times, timezones + durations

[Jan @uicoded]

After couple hours 🤷 😡
No, @boltdotnew does not run @nextjs version 15.x (with or without Turbopack).
FYI: @stackblitz official template is v13.5.1 and TW 3.3.3
@boltdotnewhelp

[Jan @uicoded]

@boltdotnew Decent upgrade, better DX than v0 github integration