theres higher demand for quality talent than there is supply of quality talent. i see so many startups hiring for 10+ roles everyday.
if you are an ambitious, curious, high-agency person, your time is now.
hit me up if you are looking for a role and ill see if i can help you out
brody retweeted
We need to bring back calling people.
Why have we normalized Calendly culture?
You want to talk to someone, so you send them a scheduling link, coordinate availability, and somehow end up speaking three weeks later.
Just pick up the phone and call them.
90 seconds. 120 seconds. That’s all you need to make an introduction, solve a problem, or move a deal forward.
We’ve created so much unnecessary bureaucracy around basic human interaction.
Call more people.
grok bot > granola connect > sweep meetings > add/update CRM > summarize action items > build tickets > triage tickets > transform into yes/no multiple choice decisions > brodus receives info > assigns execution bots > bots run into issues > build tickets > transform yes/no mul-
so this just ends in us being data labelers doesnt it
i think github for:
> tax / accounting filings
> financial models
> legal contracts
will be a necessary product for the future of knowledge work. the agents people operate will need a formal form of version control, privacy/encryption, and embedded governance on a shared cloud environment.
that substrate used to be google docs, but it doesnt mesh with agentic work. there are much better form factors for collaboration yet to arise, and ive been playing with apps in beta that make agent-to-agent comms fairly seamless.
cleanest example of this emerging work pattern was working with my accountant:
> i create crypto slop txn reconciliation off an amalgamation of data sources and forms
> *export claude project, rollup 11 forms and csvs into a zip*
> *send zip*
> acctountant receives it
> *download/unzip* *hope this isnt malware*
> *checks files* "nice, night malware!"
> accountant set his agents on my folder
> agents apply accountants frameworks / expertise to extract insights and status of my files
> accountant sends back review
> we go back and forth on messages
pretty cool, but what if:
> i create my slop reconciliation
> open a comms channel with my accountant
> send accountant an .md summary of my tax session history, processes, along with permissioned mcp end points, csvs, addtl context
> accountants agents receive request via an api, charge me $x for the review
> i ack the fee, pay
> accountants agents review
> ships to human accountants for final edits/review
> files are all versioned, uploaded to the teams repo
> accountant stamps approval, ship to IRS
in and out, trustlessly. actual agentic services and commerce. the human role and qualifications will evolve and so will the tooling to fit these different work patterns
I just launched an affiliate UGC program for my AI-enabled SMB RL env / data reselling business that I reverse-engineer vibe coded from a research loop of Micro1 on a DGX spark on-prem cluster hosting GLM 5.2 and software is so cooked I played COD zombies on the browsers while watching madison beers kick stream which I can understand why the chargers are 0-5 now, bench Herbert, and 3d print me a Glock 19 with a switch because no way I make it out of this p-doom psychosis before the tesla Optimus bots take over and our country is flooded with flankers!
x via mcp in grok bot is a big unlock for getting more out of the platform.
im now solely interfacing with my bookmarks, x posts, and accounts as if it were any other dataset that my agents have access to.
easy workflow
> see a playbook / insight
> bookmark
> grok cron pull bookmarks
> extract playbook / insight > apply it to my projects / clients / goals
> pull playbook into main for ea. project
> test / eval playbook
brody retweeted
You want to own your intelligence stack and everything you build around an AI model for two reasons
1. This is core business IP versus getting IP farmed by the labs
2. It's the only way to drive unique and differentiated outcomes versus using the generic model everyone else uses
Microsoft CEO Satya Nadella wrote that a company renting an AI model “essentially pays for intelligence twice”:
Once in money, and again in the proprietary knowledge it has to reveal to make the model useful.
This mechanism is called the learning loop: every prompt and correction captures some of the company's know-how.
That knowledge is distilled and, with a closed model, the provider keeps that information.
If your usage data is worth more than the token cost to serve, providers might end up paying you to use their models. The precedent exists: OpenAI created an opt-in program that gives accounts up to 1M free tokens a day on certain models in exchange for training rights on that traffic.
Chips can be rented and algorithms can be leaked or distilled, but the prompts and corrections flowing through a model compound privately for whoever holds them.
brody retweeted
IT’S AN IOWA HAWKEYES MIRACLE AT MICHIGAN
ANOTHER SECOND ADDED WOULDN’T SAVE MICHIGAN
Seeing a lot of crazy stuff come out around Jev and I am curious how this effects the revenue model for LLM neo clouds / inference providers.
Until now all growth assumptions were modeled off of LLM usage economics, but enter these System One models like Jev, and it drastically shifts inference assumptions from 100% sequential token processing LLM-based inference to this parallelized alternative that's clearly better for many workloads people were using LLMs for.
If LLMs were used for 100% of tasks, but x% of tasks are better for SO models, does the inference spend for the remaining y% of LLM make up for the chunk that's been zeroe'd by x% of SO tasks?
I'm reluctant to call this a neocloud revenue model nuke but it's hard to see how it's not. Pls help me understand.
Okay so in a business setting, what % of tasks are actually high entropy? I would argue that few actually are and that 99% of business cases (emphasis on business, excluding scientific / mathematical discoveries) are low-medium entropy.
xitter.cf/alexatallah/status/209…
I tell many of our customers that they should be decomposing their AI tasks into three buckets:
1. Low entropy: e.g. "who should review this PR?"
2. Medium entropy: e.g. "review this PR"
3. High entropy: e.g. "write a PR"
Today, I think you only need frontier models for #3!
hey jack, would be nice to be able to add our hermes setup 🙂
if the setup is anything easier than how i currently interface w telegram / discord gateways, ill switch rn
we're launching BUZZ!
a new groupchat platform for teams of people and agents of all sizes, built to reduce our dependency on slack and github. model-agnostic, decentralized, self-sovereign, and open source. 🐝
buzz.xyz
