Nat Taylor

Product Manager, AI tinkerer, web developer, sailor, dad and CoCo'09 grad

[Nat Taylor @nattaylor]

The third time is a charm. We made it

[Nat Taylor @nattaylor]

6hr delay, board, deplane, board, taxi, return to gate and @AmericanAir has 2 bags of pretzels to offer.

Maybe time to heed the engine's warnings and not try to push it trans-atlantic

[Nat Taylor @nattaylor]

Hope this isn't my last post. They gave it the ol' turn-it-off-turn-it-on once again and are trying again

[Google Gemma @googlegemma]

Gemma 4 just got even faster!

We're releasing Multi-Token Prediction (MTP) drafters that deliver up to a 3x speedup, without any degradation in output quality or reasoning logic.

[Nat Taylor @nattaylor]

@samokhvalov Ok 😁 challenge accepted: Here's a Codex build of the extension I've always sort of wanted https://t.co/icOCtsntK1

[Nat Taylor @nattaylor]

@samokhvalov I like it _in the database server_ 😈 Snowflake's implementation works pretty well!

[Josh Woodward @joshwoodward]

New in Gemini: Generate files and export them

Tell Gemini what you want to create and the format, and it now does the work for you.

Now supporting:
📄 Google Docs, Word (.docx) & PDFs
📊 Google Sheets, Excel (.xlsx) & CSV
🖥️ Google Slides
🛠️ Markdown, LaTeX, TXT, RTF

Available now on all surfaces globally!

[Nat Taylor @nattaylor]

@antoine_chaffin @LightOnIO I can't wait to see the speedy stuff people build with all this!

[Raphaël Sourty @raphaelsrty]

Hi, we are releasing ColGrep 1.2.0

ColGrep now incorporate BM25 trigrams to further enhance our multi-vector models using hybrid search.

Now, ColGrep print relative paths by default (fewer tokens per result)

Exact same features as GREP

Improved CUDA usage and installation

[Nat Taylor @nattaylor]

Amazing that reasoning can be compressed into the weights

[LightOn @LightOnIO]

💫 Introducing New SOTA Long Context VLM

LightOn OriOn-Qwen-SR1 reasons over full documents and executes it implicitly at inference.

Reasoning is compressed into the model's weights, no verbose output, no added latency.

🥇 SOTA on MMLongBenchDoc, ahead of Qwen3 VL with 7× fewer parameters.

🙌 Kudos to @further_ai for this new milestone!

Reasoning starts at reading.
👉 https://t.co/fLsnY9USeA

[Lucas Atkins @latkins]

This April Fool's Day, we decided to stop joking around.

Trinity-Large-Thinking is out now.

[Andrew Lamb @andrewlamb1111]

Another example of incorporating the latest Database Research (in this case Limit Pruning from ❄️) in @ApacheDataFusio
https://t.co/7fjsU5zWAR.

[Marcelo Trylesinski @marcelotryle]

Starlette 1.0 is here!🎉

After nearly eight years, Starlette has reached its first stable release. Downloaded almost 10 million times a day, it serves as the foundation for FastAPI and the Python MCP SDK.

Blog post: https://t.co/KdSmoBJnah
Release notes: https://t.co/m8Si0XDexJ

[Alex Garcia @agarcia_me]

we're back baby!

[Nik Samokhvalov @samokhvalov]

rpg 0.8.0 – modern PostgreSQL terminal https://t.co/UyLC2wNzhi

in this release:

EXPLAIN plans that actually help you

Instead of staring at a wall of text, rpg now shows a summary header with execution time, buffer stats, and automatic warnings — seq scans on large tables, row estimate errors, sorts spilling to disk. The full plan is preserved with color coding so the hot path jumps out instantly.

Plan export to popular visualization/analysis tools

When you find a slow query, share the plan in one command:
\explain share depesz
→ https://t.co/JvBpX233tE

Works with depesz, dalibo, and pgMustard. No copy-paste, no browser tabs.

Lua custom commands

Write a Lua script, drop it in ~/.config/rpg/commands/, and you have a new \backslash command.

Other highlights
- Markdown output (\pset format markdown; psql --markdown) — paste query results directly into GitHub issues
- Better \s history — filter with patterns, syntax highlighted, through the pager
- Tab completion for all rpg commands
- pspg integration with 20+ themes
- lots of fixes based on feedback I've got yesterday/today

⭐ https://t.co/volv01adzo

[Antoine Chaffin @antoine_chaffin]

Well, well, well...
One perks of building on top of sentence-transformers is that everytime @tomaarsen ships, PyLate gets free features...
And I am hearing that lately he might have focused on... multimodality
https://t.co/BRvudwrlG9
Yes, ColPali will soon be supported by PyLate :)
(Actually, not only images 🚀)

[Nathan @nathanhabib1011]

NEW SOTA OCR MODEL DROPPED
Congrats to @VikParuchuri and team for releasing Chandra OCR 2!

- 85.9% on olmocr bench, making it first place 🏆
- 90+ language support
- 4B model
- Full layout information
- Extracts + captions images and diagrams
- Strong handwriting, math, form, table support

Compare every OCR model on the hub and choose the one adapted to your needs 👇

[LightOn @LightOnIO]

The multi-vector era is here and there is no going back.
Reason-ModernColBERT tops BrowseComp-Plus, the hardest agentic search benchmark available, by 7.59 points on accuracy.

🥇on accuracy.
🥇on recall.
🥇on calibration.
📉 Fewest search calls.

The models it outperforms? Up to 54× larger.

Reasoning-intensive retrieval (BRIGHT), code search (MTEB Code), agentic Deep Research (BrowseComp-Plus). The pattern is the same: late interaction dominates, with a fraction of the parameters.

149M parameters. Open weights. Open code. Built with PyLate in a few hours.

Full results, analysis and recipe on LightOn blog: https://t.co/7PiseAwhQP

[Rohan Joshi @ron_joshi]

Introducing Kitten TTS V0.8: open-source TTS that fits in 25MB.

Three variants: 80M | 40M | 14M (<25MB)

Highly expressive. Runs on CPU. Built for edge.

No GPU? No problem. Ship voice anywhere.

Check it out:

[Antoine Chaffin @antoine_chaffin]

BrowseComp-Plus, perhaps the hardest popular deep research task, is now solved at nearly 90%...
... and all it took was a 150M model ✨
Thrilled to announce that Reason-ModernColBERT did it again and outperform all models (including models 54× bigger) on all metrics