11 Tricks to Make Your AI Application Fast Practical techniques to reduce latency and improve perceived performance in AI applications — from model selection to UI patterns.

→ Validate structured output mid-stream, fail fast
→ Buffer rapid requests into single LLM calls

Full writeup with interactive visualisations: betalyra.com/blog/11-tric...

→ Structure prompts for caching (static top, dynamic bottom)
→ Disable thinking — it's on by default, adds seconds
→ Compact conversation history every turn
→ Strip your system prompt to the bone
→ Push everything except the response to background queues

Integrated AI into your SaaS but it feels slow? 11 tricks to make it fast:

→ Use small models, route up when needed
→ Minimise output tokens — they're 2-4x more expensive
→ Stream everything, including tool call results
→ Design use-case tools, not CRUD wrappers

Tiles Changelog All notable changes and releases for Tiles.

Tiles version 0.4.3 Alpha 7 has been released.

tiles.run/download

Turn-based chats with SQLite persistence. Release artifacts now include a .pkg bundle for easier installs. Support for non-Harmony models such as qwen 3.5-4b. Bug fix for venvstack generating multiple tar archives.

Introducing ChatGPT Atlas The browser with ChatGPT built in.

OpenAI just announced their own browser Atlas.

After Perplexity’s Comet and Claude’s Chrome extension OpenAI joins the AI-native browser wars. This will put some serious pressure on Google search..

#aio #geo #ai #llms #aiso
openai.com/index/introd...

Introducing ChatGPT Atlas The browser with ChatGPT built in.

OpenAI just announced their own browser Atlas.

After Perplexity’s Comet and Claude’s Chrome extension OpenAI joins the AI-native browser wars. This will put some serious pressure on Google search..

#aio #geo #ai #llms #aiso
openai.com/index/introd...

Instead of a useless AI emoji generator I wish #apple had used AI to create intelligent printing that doesn't print the whole page black or print in portrait mode even though you selected landscape

Architecting and Evaluating an AI-First Search API Building a scalable Search API that handles 200 million daily queries using hybrid retrieval and intelligent context curation for AI models

So pretty similar to a typical RAG pipeline but on a giant scale.

The search api announcement:
www.perplexity.ai/hub/blog/int...

The insights article
research.perplexity.ai/articles/arc...

hashtag#aio hashtag#aiso hashtag#geo hashtag#ai hashtag#llm hashtag#aeo

The gist is:
- they start with semantic embeddings and lexical retrieval (probably BM25 and similar) for a first very fast retrieval
- then they apply heuristics based filtering and base filters for things like stale content
- then using cross-encoders aka rerankers to rank the final result sets

if true this would make it an amazing api for AIO research.

They also published an article giving some super interesting insights into how their index is constructed.

This is huge! Perplexity just released their search index as an api and with it a blog article giving some insights into how they rank. According to their announcement the search api is the exact same as the one powering the perplexity application.

Yes exactly, there is no need for a special treatment of schemas

So the model "knows" that these tokens belong together in this context of a json schema.
Here is an example of how that looks in practise for the prompt "What organization is that?" where you see that the model looks at the context & type

The tokens are transformed into embeddings with a positional encoding added to them (e.g. RoPE), so tokens close to each other are more "similar" than tokens further away. This way the attention mechanism learns to attend to these tokens together (being trained on thousands of schema examples).

New captcha idea: let #ai generate a heatmap scale. #claude at least is surprisingly uncapable of doing it:

Can recommend better-auth which has organization support

TANK vs TESLA

"We've crushed fascism before and we'll crush it again"

- ⁠WW2 veteran Ken, 98, in the tank

There are many good corporate blogs that are highly informative, blog.logrocket.com eg is one I have in high esteem

Portugal with full blackout and bad internet. No jira, no teams, no AWS. Life is good

When war crimes go unanswered, and handshakes with smiles replace consequences, the message is clear: kill more — we allow it.

Building Effect 4.0 | Michael Arnaldi (Effect Days 2025) YouTube video by Effect | TypeScript at Scale

Effect 4.0 is on the way – smaller, faster, and more refined. In his closing remarks, Michael Arnaldi, creator of Effect, shared what’s coming:

▸ ~20× faster streams & batching
▸ Smaller bundles
▸ STM integrated into core

Big changes, same focus.

www.youtube.com/watch?v=nyvB...

import { oc } from '@orpc/effect' ;)

awesome! would be great if you could add support for Effect- and Effect<Stream>-based handlers ;)

AI generated marketing videos saas

Want an in depth exploration of the different hardware architectures within AI?
Of course you do :)
Another great article by Chris Fleetwood:
fleetwood.dev/posts/domain...

Yeah not saying this is the case with every traditional bank and some even have fast transfers. But there are also the traditional banks where you wait 3 days for an intra-EU transfer, speaking from experience, while with my modern bank the transfer arrives in seconds.

just give me real-time indexing. It's 2025!! The technology is there. Perplexity & co can leverage this to appear always up to date in comparison to Google, Bing & Co

This feels like traditional banking vs. modern banking like revolut. Google feels like traditional banks that rely on some batch COBOL jobs that some dude in the 70s wrote and that run once a night and if it fails, will run again the night after. also not running on the weekends.

search engines newcomers like perplexity.ai, chatgpt.com, you.com outperform google with their AI, but there is another attack vector that they can leverage: indexing. Google & Bing take like ages to index a page, even if you submit sitemaps & stuff.

If the United States is willing to extort with Starlink, assume AWS, Google, and Azure clouds are next.

Ukraine today, whose turn is it tomorrow?

Any US technology is now a liability.

We need European solutions as fast as possible.