Chirag Modi

2025 motto - you can just do things!

super cool. worth checking out!! :))

@adamshuaib Spot on. That's why I took references with a pinch of salt during DD.

The founders worth betting on were often terrible employees. Many clashed with previous managers, didn’t follow instructions and quickly realised they were more effective than their boss. In the wrong environment these people are marginalised or ignored. It’s also why previous employer references for a founder don’t always show their true ability. This kind of mindset doesn’t tend to show up well on a CV. It looks like job-hopping or not being a team player. But in practice it signals someone who can handle the unstructured, highly uncertain nature of running a startup. Our data on 15,000 companies showed that founders who scored highly on emotional stability (not charisma or likability) were far more likely to build bigger companies.

@nicochristie Inverse relationship at least at early stage

Don’t start a company until you are ready to tie your physical health to your company’s health

@ryanbrewer 🤣 🤣 🤣

@modic123 Thankfully I work at a token provider so that doesn’t happen

I used to be able to sit in front of my computer for 14 hours straight and just code. Now I need to take breaks because I’m switching context so frequently

@ryanbrewer That’s reserved for midnight till 4am when you run out of tokens ;)

@modic123 I miss the flow state

@dnlkwk vps getting replaced

Every banker and PE guy knows the dread of printing and scrubbing decks and memos for hours. Final Pass legit reduces that down to 5 minutes. Using multiple AI models to ensure it's client ready.

Claude for all things knowledge work feels like it is on a very similar trajectory to what agentic coding experienced last year. I expect entire industries that rely on spreadsheets and powerpoints to begin to be transformed in the next few months.

@anshublog Congrats!

New: we got named Forbes Best Startups to Work for list, 3rd year in a row. At #47, we are ahead of some amazing companies we love and respect.

Sick idea 👏 👏 👏

Agree but the models will get there. An example of an Inflection point for workflows would be to have a set of slides (20+) that have a coherent narrative and underlying model that ties together. Sth that represents a good first draft that you can work with. This announcement is quite timely for this example. x.com/claudeai/statu…

@modic123 Think the real issue here isn’t whether or not some LBO and pitch deck can be created but whether your exact LBO and pitch deck can be created. It doesn’t really matter what the average output should look like. It needs to conform to your best work.

My prediction: by end of the year, SOTA models will be able to automate entire workflows in banking. We will go from doing simple LBOs / slides to creating IM / decks with the underlying model based on a VDR. Something that takes investment teams 40-80+ hours in total will be able to get done in 1h or less. It won’t be perfect and post training will play a big role.

MDs about to have a field day with this.

@quentin_o26 super cool. would love to chat to hear how its going! DMs open

we just have to tune the open source plugins to add our context, our strategies, what we are looking for, and it would be perfect. The xlsx product are really neat, word and pptx as well. I am personally found of html slides as the model can do more things / put more informations than on classic pptx where the oxml framework breaks a lot of times which encore anthropic to reduce the number of data available in a slide

@horiacle1 👏 👏 👏

Last week, we've released a huge library of skills that any agent can use to connect to any API on the internet. They are now available in Clawhub, Skills.sh, Smithery and many other skills marketplaces. The response has been incredible. - Daily rate of signups 4x'd vs 1 week ago & it keeps increasing - Our self-integration skill for @openclaw (lnkd.in/ePE7VeDa) has crossed 400 downloads already - Across skills per app, we have already crossed 200 installs into Claude Claude, Codex, Replit, Openclaws & other agents So what do these skills do? They give your agent capability to access any API you want and let your agent build that integration or act based on data from a specific app. It's so painful to have to go have your agent write incorrect API calls or not figure out the right Auth mechanism. This completely shortcuts it and give your personal agent (like Openclaw) or coding agent (like Claude Code) the capability to integrate on the fly. If you are interested in trying out any of the skills, just point your agent to the links below 👇

@jeffgrimes9 nicely done!!

Perplexity Finance earnings previews now include average 1-day price move (based on the last 4 earnings), and options-implied 1-day price move. Available to all users.

@RekhiVarun lol

Oil about to go vertical. Buckle up.

@ajratner spot on

Excited chatter about "RL environments" continues - but the environment itself (as often defined) is only one small part! Environments (for eval or tuning/RL) require - Tasks that reflect model failure modes + real world usage distributions - Detailed task-specific rubrics - Reference output and traces. Just like a desk with a calculator isn't much without test questions, answers, and grading keys. Depending on how you define the "environment", there are many dynamic components needed as well - e.g. simulated data, user personas for multi-turn / collaboration, etc. The "environment" itself (as often narrowly defined) - e.g. a website/CRM/etc clone - is generally the *least* important part, especially given intersection with today's agentic coding capabilities. Tl/dr: Pay most attention to the data and dynamic components - not just the "environment" itself.

@adamshuaib @gdb only in certain domains. :))

@gdb We’ve already reached the point where the models are smarter than the people writing the test questions.

Benchmarks? Where we’re going, we don’t need benchmarks.

@dustinthomas02 @TheRealAdamG of course. try uploading a ref doc in whatever tool you have access to!

@modic123 @TheRealAdamG I’ll work with the team to see if we can review those. Always red tape to get through. Appreciate the insights.

While ChatGPT for Excel is all the rage, the core presentation generation capabilities of GPT-5.4 took a huge step forward (with more room to go).