chief researcher at https://xitter.cf/t.co/85QLNI0SE9 | working at the intersection of business processes, neural network topologies & machine learning
San Francisco, CA
Joined February 2013
- Tweets2.3K
- Following312
- Followers838
- Likes1.1K
Somesh Misra / ERP.ai retweeted
Proto is now the #2 productivity agent on @OpenRouter
486B tokens today
We think it’s happening because Proto is built on one belief: you shouldn’t be locked into a single model.
A harness tied to one provider bets your whole business on one lab’s roadmap and pricing. Proto runs on any model, so you always get the best one for the job and never pay a lock-in tax to OpenAI or Anthropic.
An agent inside Proto isn’t a loose cannon with a god-mode key that might do anything random. It works with the permissions of a real enterprise user.
It can see only what its role can see, do only what its role can do, and every action goes into the same audit trail as a human’s.
That’s the difference between an AI demo and an agent a company can actually hand its operations to. Any model, real permissions, full accountability. That’s Proto.
Somesh Misra / ERP.ai retweeted
ERP.AI launches today as headless SaaS: one business core where people, agents, and integrations all operate the same apps, records, and workflows through APIs (with the right permissions).
Proto, their desktop agent, builds the custom app from a plain description of a process you know (invoices, purchasing, etc.), then works the records - finding overdue items, drafting follow-ups, applying rules - while keeping human approval on anything that sends or pays.
Screens become optional. Agents become the operators.
🤝 Paid partnership
For 25 years, Salesforce, SAP and enterprise software were built around one assumption: humans operate the software.
We’re changing that.
Introducing ERP•AI and Proto: business software built for agents, and an agent that operates business software.
Describe the system you need. ERP•AI builds it.
Then Proto takes over. It manages records, runs workflows and executes actions inside the software you just created.
This is Headless SaaS.
1: get ERP•AI
2: build your SaaS
3: post below
Best one wins a MacBook. First 100 decent entries get $100.
(a message from Steve Papa / founding team)
Somesh Misra / ERP.ai retweeted
Things are getting crazy: This platform lets you build an app and use it to do your work.
It's both the Builder and the Operator.
1. You want to build an invoicing app, so you describe it to Proto and specify the fields you want to track.
2. Proto turns your description into the invoicing app. It creates the data structures, views, and workflows you need.
3. You verify the invoicing app and make sure everything works.
4. From here, you can tell Proto to start invoicing your clients by using the app it just built.
For 25 years, Salesforce, SAP and enterprise software were built around one assumption: humans operate the software.
We’re changing that.
Introducing ERP•AI and Proto: business software built for agents, and an agent that operates business software.
Describe the system you need. ERP•AI builds it.
Then Proto takes over. It manages records, runs workflows and executes actions inside the software you just created.
This is Headless SaaS.
1: get ERP•AI
2: build your SaaS
3: post below
Best one wins a MacBook. First 100 decent entries get $100.
(a message from Steve Papa / founding team)
Somesh Misra / ERP.ai retweeted
I THINK ERP.AI JUST SHOWED WHAT BUSINESS SOFTWARE IS GOING TO LOOK LIKE, AND THIS CLINIC EXAMPLE EXPLAINS IT PERFECTLY.
A small clinic running on 4 tools, 4 logins and $642 a month.
None of them talk to each other.
Then one prompt to Proto: "Build our clinic front desk."
Prompt → Calendar → Intake Forms → Roles → Waitlist Agent.
→ Appointments, intake forms and reminders end up in one app.
→ Every role only sees what it should.
→ A patient cancels the 3 PM slot and the agent texts the first person on the waitlist.
→ They reply YES, get booked and receive their intake form.
→ The slot gets filled in 3 minutes without anyone touching it.
You describe how your business works. It builds it and then runs it.
For 25 years, Salesforce, SAP and enterprise software were built around one assumption: humans operate the software.
We’re changing that.
Introducing ERP•AI and Proto: business software built for agents, and an agent that operates business software.
Describe the system you need. ERP•AI builds it.
Then Proto takes over. It manages records, runs workflows and executes actions inside the software you just created.
This is Headless SaaS.
1: get ERP•AI
2: build your SaaS
3: post below
Best one wins a MacBook. First 100 decent entries get $100.
(a message from Steve Papa / founding team)
Quite a bit happening under the hood, including some interesting use of GNNs for business processes.
For 25 years, Salesforce, SAP and enterprise software were built around one assumption: humans operate the software.
We’re changing that.
Introducing ERP•AI and Proto: business software built for agents, and an agent that operates business software.
Describe the system you need. ERP•AI builds it.
Then Proto takes over. It manages records, runs workflows and executes actions inside the software you just created.
This is Headless SaaS.
1: get ERP•AI
2: build your SaaS
3: post below
Best one wins a MacBook. First 100 decent entries get $100.
(a message from Steve Papa / founding team)
Somesh Misra / ERP.ai retweeted
Really excited about this project - we've been working on it in stealth for 5 years! Re-imagining how business applications are delivered from the enterprise through SMB. A true revolution is starting here....
For 25 years, Salesforce, SAP and enterprise software were built around one assumption: humans operate the software.
We’re changing that.
Introducing ERP•AI and Proto: business software built for agents, and an agent that operates business software.
Describe the system you need. ERP•AI builds it.
Then Proto takes over. It manages records, runs workflows and executes actions inside the software you just created.
This is Headless SaaS.
1: get ERP•AI
2: build your SaaS
3: post below
Best one wins a MacBook. First 100 decent entries get $100.
(a message from Steve Papa / founding team)
Somesh Misra / ERP.ai retweeted
For 25 years, Salesforce, SAP and enterprise software were built around one assumption: humans operate the software.
We’re changing that.
Introducing ERP•AI and Proto: business software built for agents, and an agent that operates business software.
Describe the system you need. ERP•AI builds it.
Then Proto takes over. It manages records, runs workflows and executes actions inside the software you just created.
This is Headless SaaS.
1: get ERP•AI
2: build your SaaS
3: post below
Best one wins a MacBook. First 100 decent entries get $100.
(a message from Steve Papa / founding team)
Somesh Misra / ERP.ai retweeted
Steve Papa is in a league of his own. Easily one of the best entrepreneurs of our gen.
- built Endeca to a $1.1B exit in 2011 (back when billion dollar exits were very rare)
- incubated Toast (founders were ex Endeca PMs)
- took on the incredibly complex telco industry with Parallel Wireless
His range is extraordinary.
Give him a follow @stevepapa20
722 new AI generated math papers in one day. many people are reading this as the end of mathematicians 😅. I read it the other way: this is raw material. Mathematicians can now build new theories more easily than ever, if we treat these proofs as starting points, not endpoints.
We have already seen this happen. In April, AI proved Erdos 1196. Lichtman's team took the key idea and used it to prove a 60 yr old Erdos Sarkozy szemeredi conjecture too. one AI proof became a new method.
x.com/jdlichtman/status/2050…
🧵
We’re releasing a broad range of new mathematical results produced by an internal frontier model.
We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.
github.com/openai/math
But that only happened because humans understood the proof. A proof tells you something is true. Understanding tells you why, and the why is what you can reuse on the next problem/theory. Without it, a proof is a dead end.
Take Navier Stokes. OpenAI's proof is believed to be correct, but mathematicians say it is very hard to read and they have learned little from it so far. Being correct is not the same as being useful. Until someone understands it, it's a locked door.
So the bottleneck is moving. Finding proofs is getting cheap. reading them, explaining them, and turning them into ideas a person can hold in their head is now the scarce work. As Daniel Spielman has argued, that work deserves real credit.
For those submitting to AI capabilities without understanding the math,
AI won't end mathematics 🤣
It will move it up a level. The machine finds the paths. Mathematicians draw the map, name the ideas, and build the next layer of abstraction on top. That job just got bigger, not smaller. @nasqret
Proto has a graph-theoretic backend, because that's what a business actually is.
This is exactly what I meant last month!
as more of the work of formalizing and checking proofs becomes routine, mathematicians will climb higher, towards new abstractions, objects and conjectures. Perhaps that’s where the genius goes!
Fermat's Last Theorem is finally sealed. Congratulations to all the contributors, especially Kevin Buzzard and his vast team of students and post-docs who did the big part of the proof in the project
github.com/leanprover-commun…
But the story continues because the proof is not yet formalized in Heyting arithmetic. This will be the final bit to show that the proof is really elementary (though a bit long...). It will be a giant lambda function with the right type. Shall we try? I started building Heyting arithmetic library last month and so far it goes great
bnaskrecki.faculty.wmi.amu.e…
Somesh Misra / ERP.ai retweeted
Fermat's Last Theorem is finally sealed. Congratulations to all the contributors, especially Kevin Buzzard and his vast team of students and post-docs who did the big part of the proof in the project
github.com/leanprover-commun…
But the story continues because the proof is not yet formalized in Heyting arithmetic. This will be the final bit to show that the proof is really elementary (though a bit long...). It will be a giant lambda function with the right type. Shall we try? I started building Heyting arithmetic library last month and so far it goes great
bnaskrecki.faculty.wmi.amu.e…
Checking that a major mathematical proof is correct can take years. Formalization—converting the mathematical reasoning into a form computer proof assistants like Lean can verify—can help.
Last month, Claude completed the first formalized proof of Fermat’s Last Theorem, one of the most famous theorems of all time. This was a project experts thought would take many years. It is the largest Lean proof ever written.
Fermat’s Last Theorem was first proven in 1995 by Sir Andrew Wiles, more than 350 years after it was conjectured. Our proof, which totals over 13 million lines of code, provides machine verification. More importantly, it proves over 29,000 other theorems that the proof requires, across many areas of math which had never before been formalized.
We see this as a major step in the long process of firming up the core of mathematical knowledge, building on work from three centuries of mathematicians and hundreds of contributors to Lean and Mathlib. We are optimistic that AI-assisted verification of mathematical proofs will help reduce the burden of refereeing mathematics in an era where more proofs are being produced than ever before.
You can read about the process on our Science Blog: anthropic.com/research/forma…
And see the complete proof on GitHub: github.com/anthropics/fermat…
