|
Varsity Women's Club
|
Issue 09 · July 2026
|
|
|
Issue 09
The Open Playbook Issue
Thinking Machines, an Open AI spin out, just showed its cards, and its whole strategy is a bet that custom training beats raw talent. Sound familiar? Plus, nine open roles at Nowadays are still live.
|
|
⚡ The Scouting Report
A Rookie Lab Publishes Its Playbook
|
|
|
Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, released its first model on Wednesday. It's called Inkling, and the headline isn't how smart it is. It's how it's being handed out.
Inkling is open-weight, which means any company can download it, look under the hood, and retrain it on their own expertise. Compare that to ChatGPT, Claude, and Gemini, where you rent access to a model someone else controls, and you can see why this is being called a bet against one-size-fits-all AI.
|
|
975B Parameters
But it only activates about 41 billion per task, which keeps it fast and cheap to run.
|
|
45T Tokens
Trained on 45 trillion tokens of text, image, audio, and video.
|
|
1/3 the Tokens
On one coding benchmark, it matches Nvidia's Nemotron 3 Ultra while using a third of the tokens.
|
|
9 Months to Ship
Thinking Machines shipped a market-ready model in about nine months, a pace that took OpenAI roughly five years and Anthropic roughly three.
|
|
|
Here's the part that should get your attention. Thinking Machines openly admits Inkling is not the strongest model on the market. Their bet is that it doesn't need to be, because a solid model trained on your organization's specific expertise beats a superstar generalist. The proof point: researchers fine-tuned an open model on Bridgewater Associates' financial knowledge and beat top proprietary models on financial reasoning tests, at roughly one fourteenth the running cost.
Even Microsoft CEO Satya Nadella is sounding the alarm on the other side of this trade, warning that enterprises using proprietary models pay twice: once in subscriptions, and again by handing over their business knowledge through every prompt and correction. Having a model custom trained on your data that you own keeps trade secrets in house.
|
|
A solid model trained on your organization's specific expertise beats a superstar generalist.
|
|
|
Why It Matters for You
Athletes know that raw talent is just the baseline. The real wins come from focused, disciplined training. The AI industry is realizing the same thing. Companies are looking for professionals who can take a general-purpose model and refine it into a high-performance specialist. If you're building skills in fine-tuning, prompt engineering, or agent design, you're training for exactly the type of work companies need right now.
|
|
|
Read the Full Story →
|
|
|
|
★ This Monday · Agent Camp
Guest Coach: A Forward Deployed Engineer
This Monday · Agent Camp
We're bringing in a guest who does this for a living. She's a Forward Deployed Engineer, an AI agent expert whose entire job is building high-functioning, accurate agents for real customers, then standing next to them while the agents run in the wild.
If you've never heard the title, here's the short version: a forward deployed engineer is the person a company sends into the field to make AI actually work. Not in a demo. Not in a slide deck. Inside a customer's messy, real-world operations, where the data is imperfect and the stakes are live. It's one of the fastest-growing and hardest-to-fill roles in tech right now, and it rewards exactly what athletes bring: reading the field in real time, adjusting on the fly, and owning the outcome.
She'll walk us through what her day actually looks like, how she designs agents that stay accurate under pressure, and what separates an agent that demos well from one that survives contact with reality. Enrolled in Agent Camp? Come with questions. Not enrolled yet? Reply to this email and we'll tell you about the next cohort.
Save Your Seat →
|
|
|
Now Hiring
Stealth CPG Healthtech Startup
|
|
Supply Chain Lead
NYC (remote-friendly) · Health Tech CPG
A stealth health tech CPG company that grew from $0 to $100M ARR in 14 months is searching for a Supply Chain Lead. The role is NYC-based and remote-friendly.
|
|
Nine Ways Into Nowadays
|
|
Nowadays (YC S23)
San Francisco · AI Event Platform
Anna's team builds for Fortune 500s, for fellow YC startups, and for themselves. Their own offsites set the bar: the team races Ironmans and marathons together, and one retreat took them to the Thailand resort from White Lotus season 3. If planning events and automating the messy parts with AI sounds like your arena, this is a team worth meeting.
|
|
|
Engineering
Nowadays · San Francisco
|
|
|
Founding engineering is a two-way sprint: architect for the long game, ship for today.
|
|
|
Founding Software Engineer · full stack, from AI systems to React →
Founding Product Engineer · customer-driven features, end to end →
Founding AI Engineer · LLM agents for calling, scraping, research →
Product Engineer Intern · Fall 2026 →
|
|
Operations
Nowadays · San Francisco
|
|
|
Ops is the engine room: own events end to end and turn chaos into repeatable plays.
|
|
|
Founding Operator · own events and client comms end to end →
Operations Intern · Fall 2026, RFP to proposal →
|
|
Design
Nowadays · San Francisco
|
|
|
Own the whole system: brand, product, and the voice everything speaks in.
|
|
|
Founding Product Designer · design system and brand from scratch →
|
|
Go-To-Market
Nowadays · San Francisco
|
|
|
Full-cycle selling with founder-close energy: own the conversation from first call to signed deal.
|
|
|
Founding Account Executive · full sales cycle, first call to close →
Founding Mid-Market AE · deals from $50K to $200K ACV →
|
|
How to Apply
Interested in any Nowadays role? Just reply to this email, or write [email protected], and we'll get you on a call with the team.
|
|
|
Grow the Roster
Know a teammate who belongs here? Forward this, and send them our way.
Share VWC →
|
|
|
|
