shafin

Introducing Trouve - the browser agent for complicated, legacy enterprise applications such as SAP, Oracle’s NetSuite and ServiceNow. On NetSuite, Trouve can complete ANY task in under 4 minutes. It is up to 12x faster than both Claude Chrome with Opus 4.7 and Codex with GPT 5.5. We have customers who've stopped clicking through NetSuite entirely. They just prompt Trouve instead and they’re doing that non stop for up to 6 hours a day! Trouve uses best practices of the software, does not go into guessing loops like general purpose browser agents, and can adapt to custom implementations of enterprise software.

can’t help but fanboy over codex via phone holy shit

given that codex's official stance has always been "yes of course please use the app server for any frontend you want" and claude's has been "uhhh not sure if doing that with your sub allowed but we'll get back to you on that" i can't say i'm too surprised

i came to san francisco for one reason. to tell the dopest stories in the world:

@metapreston it is exactly why. i used suno to generate background music for my demo videos. I'm guessing people are doing the same for their corporate vlogs/content

This is my theory on why Suno is successful despite producing terrible music

@wordgrammer x.com/_shafinsiddiqu…

So are you guys still using the chat completions API for your gpt wrappers or are you just running codex on the server

@faizan10114 lmao

I am so tired of this narrative, people in Canada are not behind. We use Github copilot for coding like everyone else.

@tolstoybb which 30 year old software

Initially thought I’d use AI to make my own software but it’s actually incredibly helpful for utilizing 30 year old technical software.

This is the way. From pitch decks to sales proposals to contracts to user documentation, I’ve done everything over the last 3 months this way.

why is every article worth reading under a paywall, am i supposed to have subscriptions to WSJ, The Atlantic, The New Yorker, NyTimes, the FP, along with a whole bunch of substacks

@DhravyaShah @claudeai @densumesh I can’t believe it won by only 1 point

Filesystems vs Vector search is the new MCP vs CLI. @claudeai uses agentic search. And @densumesh at mintlify like filesystems too. but Retrieval is still used by Notion, Cursor, and others. We hit the cafes of SF to see what people want. - The debate is hot. Filesystems won by 1 point - filesystems feel very simple and intuitive. - RAG requires embedding, vector search and other stuff filesystems won. Introducing SMFS - Supermemory Filesystem we brought the best of these worlds into one single product - SMFS.ai. it's a filesystem, but agent can also do semantic search using grep. live today. Try it! Works with any sandbox, you mount and sync with cloud, it has a sync engine built in, all filetypes supported - even images and videos can be grepped. So, what do you choose - Filesystems, or vector search? we have both!

@AlpinDale elite ball knower

The type of software you write when you can't invert a binary tree

@krishnanrohit amazingly calm, jobs would have been calling every member of the board and cussing out their bloodline

This entire exchange makes me think way way better of Sam, as should you

the list goes on. the superbowl ad by anthropic was a colossally bad idea, the constant 'every one is going lose their jobs' is another one

the harsh reality is that @OpenAI just makes better decisions than @AnthropicAI. 1) Using Nvidia GPUs instead of TPUs 2) open sourcing codex means every embedded device in the world will use it someday 3) rust for codex to prepare for that world

there’s going to be a class of startups called “codex wrappers”. people just connecting codex to a bunch of MCPs and selling it as a packaged solution

@rayansadri also just demonstrates a pure lack of originality, i feel like by now there has been a 100 launch videos with either the social network music playing or just plain recreating one of the scenes from the movie

Every time I hear a song from The Social Network in a founder’s demo, I know with 99.9% certainty there’s almost no product, no moat, no tech, no nothing.Just a shiny, cinematic $20K demo built on a shitty tech stack that doesn’t actually exist and the founder thinks he’s 18 yo zuck in Kirkland House. Almost always true.

@rjchint lmao well when u put it like that

Dynamic and generative UI's make as much sense as a restaurant that changes its name and menu every week

is there any point in building your own harness if you're an agentic AI company, codex is general purpose and open source, why would one not just use that underneath? am i missing something?