Anguel

CTO @ClideyOfficial, Harvard dropout? catch me doing random things, building WhoDB

[Refresh @refresh_dev]

Introducing Blender Bench v1 (BBv1): a benchmark for frontier LLMs on 20 realistic 3D production workflows including animation, modeling, rigging, cloth & surfacing.

GPT-6 Astra leads overall; Claude Opus 5.5 wins modeling & cloth simulation.

[Anguel @anguel_hh]

what's your reply rate on linkedin

[Anguel @anguel_hh]

where can we get this

[vx-underground @vxunderground]

this goes unbelievably hard wtf

[Anguel @anguel_hh]

Feels like @github should be refunding this month’s subscription charges or giving credit with all the outages

[Anguel @anguel_hh]

@SakshiSugandhi I'm old okay

[Anguel @anguel_hh]

Let's gooooooooooooo!

[Clidey @ClideyOfficial]

We’ve heard the same problem from teams across industries — Retail, Manufacturing, Finance, Healthcare, Construction, Technology: you have more data than ever, but getting a clear answer still takes too much time. Your customer data is in the CRM. Sales data is in the ERP. Product usage sits somewhere else. Support has its own system. Then there are spreadsheets, dashboards, warehouses and a dozen other tools. The data exists, but the context is scattered.

So when a simple question comes up, the process still looks like this: ask someone, wait, pull a report, export a CSV, join data, open another tool, ask another team, and eventually make a decision.
We built WhoDB to change that.

Today, we’re making WhoDB available for free so teams can start bringing their data and decision-making together without waiting for a long enterprise rollout.
Here’s what WhoDB lets you do:
→ Connect your data. Bring together the systems where your business data already lives, without forcing migrations.
→ Build the data layer. Move and transform data so it can actually be used across your business.
→ Model your business. Turn scattered data into the things your teams actually work with: customers, orders, accounts, tickets, invoices, products, incidents and more.
→ Govern access. Apply permissions and policies across both people and AI agents.
→ Build on top of it. Create operational workflows using the same business context.

And once WhoDB understands how your business works, you can ask questions that go beyond individual databases:
Which customers are at risk of churning?
Which failed payments are linked to support issues?
Which deployment caused errors for paying customers?
What should we do next?

The goal is simple: less time finding and joining data, more time making decisions.
WhoDB is now free to sign up and try.
If your company has data everywhere but decisions still feel slow, we built this for you.
Try WhoDB → https://t.co/B7D0etn5Um

[Arslan @_ai_jedi]

I've tasted the 9-to-5. I've also tasted being a founder.

I highly recommend being born into generational wealth.

[Anguel @anguel_hh]

the kind of email subject that goes straight to the spam filter

[Anguel @anguel_hh]

took this long but experienced my first ever "oh no GitHub is down and it's affecting my work!"

good luck to the engineers hope they can sort it

[Seb Johnson @SebJohnsonUK]

A London AI Lab has just released an agent that outperforms Claude Opus 4.8 and GPT-5.5 on the task of replicating research.

NICE

The company, founded by DeepMinders, just emerged from stealth with a $50m seed led by Index.

The lab is called Inherent, and it’s just released Faraday - a 27B-parameter “AI Scientist” agent that outperforms Claude Opus 4.8 and GPT-5.5 on the task of replicating research.

It’s been trained via long-horizon RL with coding agents as a tool to learn the skills of a scientist.

This is VERY COOL. London AI is flying right now.

[ℏεsam @Hesamation]

DeepSeek published their harness. 24K stars already. it’s a React web app with everything designed to be a plugin: from agent loops, models, tools, to even sessions.

what does it mean? HIGH MODULARITY. you can swap any component with another one rather than edit the source code itself.

so it’s more of a harness garage where you can plug in your custom parts and enjoy.

the community work on this is gonna be so interesting.

[Lazy IT Guy @lazy_IT_guy]

Our CFO noticed we spend $65K a month on a proprietary cybersecurity platform called Sentinel Protocol.

She scheduled a meeting with me to discuss our enterprise software expenditures.

She wanted to know why our licensing fees had increased by 15% year over year.

I told her Sentinel Protocol is the only thing standing between our corporate data and a highly coordinated North Korean ransomware syndicate.

I explained that it uses a polymorphic encryption manifold to dynamically scrub our inbound packets.

She asked if she could get a Zoom meeting with their customer success team to discuss a volume discount.

I told her I'd set it up immediately.

The problem is that Sentinel Protocol doesn't exist.

It's an LLC I registered in Delaware 4 years ago.

I'm Sentinel Protocol.

My entire cybersecurity platform is just a batch script I wrote in 2018 that restarts our routers every Sunday at 3 AM.

The script took me 14 minutes to write.

I realized very early in my career that executives will never question a line item if you use the word quantum or heuristic.

But now the CFO wanted to speak to the vendor.

I went on Fiverr and hired a struggling voice actor from Chicago for $45.

I told him his character was a high-powered Silicon Valley tech bro who just drank 3 Celsius energy drinks.

I wrote him a 4-page script composed entirely of weaponized IT buzzwords.

He showed up to the Zoom call wearing a Patagonia fleece vest and introduced himself as the VP of Customer Success.

He was absolutely phenomenal.

When the CFO asked about lowering the cost, he scoffed.

He told her that downgrading our service tier would expose our legacy endpoints to asynchronous tunneling attacks.

He leaned into the webcam and asked if she was prepared to explain a catastrophic data breach to the shareholders.

She apologized for questioning the infrastructure.

Before the call ended, my Fiverr actor successfully up-sold her on a $12K biometric redundancy package.

I don't even know what a biometric redundancy package is.

I tipped the actor $100 and told him I'd need him again next quarter.

The CFO sent me an email thanking me for managing such a crucial vendor relationship.

I just used the new budget increase to buy a pontoon boat.

[Anguel @anguel_hh]

The interesting part for me is using local inference less as a chatbot and more as a routing and interpretation layer: figuring out whether a request is about files, mail, calendar, browser search, or a local integration, then extracting the relevant parameters and getting the user to the right place with fewer hops.

The launcher is built with Swift/AppKit, with a local Node service bound only to 127.0.0.1. It can use Ollama or LM Studio locally, or a user's own hosted-model key.

I also try to keep reading and acting separate: things like search, previews, and draft preparation can happen directly, while actions such as sending a message or creating an event require explicit approval.

Would love feedback, especially on the interaction model and what integrations would make this useful in practice.

[Anguel @anguel_hh]

Does anyone know why this would get flagged by Hackernews?

Text in comments

[Anguel @anguel_hh]

A few examples:

- “Schedule a 30-minute meeting with Raj next Friday” prepares a calendar draft for review.
- “I'm taking my partner to Las Vegas next weekend; find Airbnbs” extracts the destination and dates, then opens a curated Airbnb search.
- "Message Sam that I’m running 10 minutes late” finds the local WhatsApp conversation (using OpenWA), prepares a draft, and shows it for review before sending.
- “Find the Q2 planning deck” searches local files using Spotlight metadata and local ranking. Once found, the file can be previewed, opened, revealed in Finder, or dragged into another app.
- “Find the API pod that is restarting and show me its recent logs” can route into a read-only K9s integration and narrow the search down to the relevant pod and logs.

[Anguel @anguel_hh]

I'm building Habibi, a local-first command center for macOS. The idea is to talk to one launcher and have it route the request to the right local app or integration, instead of manually context-switching between apps.

[Anguel @anguel_hh]

Do I have a choice

[H @hemangkandwal]

New collaborated venture? I am sure we can divide and conquer. @anguel_hh you in?

[Hubert Thieblot @hthieblot]

Being a founder sounds romantic until you’re 2 years in with no traction, no money, and a family that thinks you need a “real job.”

[Anguel @anguel_hh]

ouch

[Kr$na @krishdotdev]

Pasted this into Claude’s memory and now i’m getting 10x better results by opus 5.