Introducing ShortcutXL - your favorite spreadsheet agent is now available as a desktop generalist.
Unlock Excel capabilities that no other spreadsheet agents can touch (VBA macros, What-if analyses, and more).
Have a swarm of Shortcut agents work on dozens of models concurrently / in parallel.
This is just a taste. try ShortcutXL now: https://t.co/nUHTsXO64p
I’m not naturally an X person.
I spend most of my time building, not posting. But for the next 30 days, I want to run a small experiment and share more of the raw reality of what we are learning while building at @Fundamental .
To start, here is one of my biggest realizations,
Training agents is a very different game from training chatbots.
Chatbots produce responses. Agents take actions. The difference sounds nuance, but it changes almost everything.
An agent has to interact with environments. The environments may be unstable or not designed around the model’s reasoning flow. The agent has to decide what to do, use available tools, interpret feedback, and recover from mistakes.
A frontier model can sometimes brute force its way through messy environments. But for a specialized model to become good, it needs training infrastructure that helps it understand the environments and practice inside it, so its reasoning gradually adapts to the environments.
For people building agents, what is the hardest part right now? Data, evals, environments, longterm reliability, or something else?
I'm rebuilding Excel
Excel was designed for floppy disks and fax machines, we deserve something better now
Run xlsx headless, API for agents + UX for humans, colab, embeddable
Mog is the new way to Excel
Free, Open source, Building in public: https://t.co/5KXgG1XQZG
I had early access to Opus 4.8 to benchmark on @tryshortcutai and the results have been incredible.
I had my gripes with 4.7 -- in fact I kept 4.6 as the default for spreadsheet work -- but Opus 4.8 is smarter/faster and has real Excel taste.
Full eval dump on the way
For the second time now, @tryshortcutai was independently ranked as the top Excel AI agent in the world.
We specifically build for the top analysts, and this matches what we see on the ground.
It's firmly us, then Claude, or your firm forces Copilot upon you.
@TechFundies why wouldn't you use Claude for Excel (or better spreadsheet agents *wink wink*)? The limitation here is openpyxl. You're much better off directly using an agent that is special-built for excel work
GPT-5.5 is pretty incredible at working in spreadsheets! And this is the worst it will ever be.
If you work in spreadsheets all day, see how much time it can save you in @tryshortcutai, or use it in our ChatGPT for Excel plugin: https://t.co/FAJnAO4eS7
We have been testing GPT 5.5 on the hardest spreadsheet tasks in the world (100k-1M+ cell complex models).
It is the Pareto frontier for spreadsheets -- SOTA accuracy, the fastest and the most efficient public model across effort levels.
OAI really cooked here
New onboarding flow for Shortcut. Always surprised how much training/onboarding is really needed.
Took about 2h to build, definitely feels like Opus 4.7 is a much stronger design eng than 4.6.
Introducing ShortcutXL - your favorite spreadsheet agent is now available as a desktop generalist.
Unlock Excel capabilities that no other spreadsheet agents can touch (VBA macros, What-if analyses, and more).
Have a swarm of Shortcut agents work on dozens of models concurrently / in parallel.
This is just a taste. try ShortcutXL now: https://t.co/nUHTsXO64p