Bots & Term Sheets for Bots

As a software engineer embedded in a VC, my daily routine swings from debugging code to dissecting pitch decks. It’s a weird hybrid job and I’ll take any edge I can get. Lately, that edge is a motley crew of AI tools. I’ve basically got an army of AI sidekicks helping me ship code, sift through startup hype, and keep my sanity. Meet the AI Crew (My Virtual Team)

GitHub Copilot: My tireless pair programmer that autocompletes my code while I sip cold brew. It turns my half-formed thoughts into code suggestions, from boilerplate CRUD functions to regex I’d rather not write.

Cline: An open-source coding agent that not only suggests code but can execute commands and run tests. Think of Cline as a junior dev who can follow instructions really fast – handy for automating those “did we update all the config files?” tasks. Great for spinning up quick prototypes!

Devin / Codex : Branded as “AI software engineer,” these tools are like an autonomous intern who can take a high-level spec and try to build it. It’s ambitious – sometimes more enthusiastic than accurate – but great for offloading a first draft of code or crunching through a boring setup while I focus on harder problems. Great for kicking off tasks before the next pitch meeting.

ChatGPT: The Swiss Army knife of my toolkit. Need a quick explanation of homomorphic encryption for a confused partner? ChatGPT’s on it. Brainstorming technical questions to ask an AI startup founder? It’s my sounding board. Refactoring a tricky algorithm or just rubber-ducking through a problem? Need to DeepResearch a new industry? ChatGPT never gets tired of my rambles.

Notion (+Granola): My writing and note-taking co-pilot. It summarizes meeting notes, cleans up my stream-of-consciousness ramblings, and even drafts sections of investment memos. Basically, it’s the ghostwriter for everything from deal write-ups to the witty email I send with code updates to the team. Granola is a great addition that pays attention for you when your brain has been bludgeoned by endless meetings.

Others: The lineup keeps growing – new AI assistants for design, data crunching, slide-making – you name it. If it promises to automate a slice of my workflow, I’m probably beta-testing it. (Yes, my laptop is a laboratory of AI experiments. No, I haven’t accidentally created Skynet… yet.)

Everyone talks about AI, but no one has the time to learn it. So, we found the easiest way to learn AI in as little time as possible: The Rundown AI.

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses – tailored to your needs.

Sign up to start learning.

My mornings often start with code. Whether I’m building an internal tool for the firm or prototyping a feature for a portfolio company, GitHub Copilot & Cline are there from the first line. It’s like having an eager intern pair programmer who writes the obvious stuff so I can focus on the tricky parts. Boilerplate setup, repetitive unit tests, even that one-off Python script to migrate data – these tools knock them out in seconds. I’ll sketch a function name or a comment, and bam! The LLM suggests half the implementation.

For example, I was spinning up a quick web dashboard for our partners to track portfolio metrics. I typed // fetch latest KPIs for each startup in JavaScript, and Copilot instantly offered up a loop with API calls and JSON parsing. It wasn’t perfect on the first try (I may have slightly cursed when it named a variable data five times), but tweaking its suggestion was a heck of a lot faster than writing from scratch. It’s like I write the outline and Copilot fills in the middle – sometimes like Mad Libs, but usually on point.

As my day picks up pace, I sometimes hand off more complex chores to Codex and Devin, the autonomous coding agents. Picture this: I’ve got a backlog of improvements for our internal “Augmented VC” platform (yes, even VCs need decent CI/CD and analytics dashboards). Rather than tackle them one by one, I fire up a task in Codex. Codex can break down a task (say, integrating a new API for market data) into steps, then in “Code” mode it starts coding and even running tests. It’s surreal to watch – like a robot intern typing away. I keep a close eye (I’m still the senior in the loop, after all), but Codex handles a lot of rote work. It might set up the HTTP client, call the endpoints, stub out a data model, and only ping me if it’s truly stumped. It’s a bit like having a super-speed junior dev who occasionally needs guidance but never needs a lunch break.

Then there’s Devin, the self-proclaimed “AI software engineer.” When I’m really juggling tasks, Devin is my attempt at being in two places at once. For instance, a portfolio company needed a simple mobile app to demo an idea – something our team didn’t have time to fully build. I gave Devin a high-level spec (in plain language) and let it loose. Devin went off to the cloud (where it runs with more resources and tools at its disposal) and started coding the app end-to-end: setting up a project structure, writing code for the UI, even commenting its logic. I won’t lie, the first pass felt like a very eager junior dev – it worked, but I had to refactor chunks to meet our quality bar. Still, in an afternoon Devin produced what might have taken a human a day or two. It’s the difference between starting with a rough MVP versus a blank page.

In VC, speed can make the difference in winning a deal or helping a founder under the gun, so having an “autonomous dev” to delegate to is a superpower (with a few caveats). I just make sure to review Devin’s output carefully – trust, but verify (the same rule I apply to overly optimistic founders!).

If Copilot and crew handle the code, ChatGPT handles the everything else. This AI has effectively become my on-demand analyst and tutor rolled into one. Need to get up to speed on a new domain? I’ll prompt ChatGPT: “Explain the core concept of quantum-resistant encryption in simple terms,” so I can follow a pitch from a crypto startup without my eyes glazing over. Preparing for a meeting with an AI biotech founder? I ask, “What are smart questions a VC engineer should ask a machine-learning drug discovery startup?” – ChatGPT will spit out a helpful list that makes me sound like I spent all weekend reading up on the subject (when in reality I was busy debugging a server and maybe squeezing in a Netflix episode or two).

ChatGPT’s also my brainstorming partner. When we’re exploring, say, the latest craze in generative AI, I’ll have a back-and-forth with ChatGPT to pressure-test ideas. I can debate it on technical merits (“What are potential scaling issues if this startup’s model usage grows 100x?”) or even role-play scenarios (“Act as a skeptical CTO and grill me on our AI platform’s design weaknesses”). It’s amazing (and slightly eerie) how it can simulate a tough conversation, preparing me for the real one.

And of course, there’s the mundane stuff: writing emails and docs. If I need to draft a friendly-but-firm email to a founder whose product isn’t ready, I might ask ChatGPT for a professional tone. It’s like having a PR assistant proofread my phrasing.

The key with ChatGPT is knowing its limits. It can hallucinate facts out of thin air, so I treat its outputs as a starting point. It’s like a well-read, confident colleague who sometimes gets the details wrong – invaluable for first drafts and idea generation, but I always double-check the critical stuff. Still, more often than not, ChatGPT has saved me from staring blankly at a page or frantically Googling for an explanation. It’s the closest thing I have to a 24/7 consultant on call, with expertise in just about every field (and ~ reasonable monthly fees, thankfully).

Working in VC means meetings, meetings, meetings – founder pitches, partner sync-ups, portfolio stand-ups – an endless parade of Zoom calls and coffee chats. Keeping track of everything is half the battle. Enter Notion. I use Notion for all my note-taking and knowledge bases, and with Notion’s AI features, I’ve basically got a second brain transcribing and summarizing on demand.

After a long meeting where a founder unloads technical details or a portfolio CTO walks through their firefighting efforts, I’ll jot down messy notes. Later, I highlight them in Notion and ask the AI to “summarize key points and action items.” The result? A clean, concise recap: e.g., “Key points: Need to hire a DevOps engineer; API response times are 300ms too slow; considering a pivot to freemium model. Actions: Introduce them to DevOps candidate Jane D.; help review caching layer code next week.” It’s not magic – I often tweak the summary – but it’s a huge time saver. Before, I’d dread turning my scribbles into something coherent (my handwriting could be a CAPTCHA test). Now it’s largely automated.

NotionAI also helps when I’m drafting longer form content, like an internal memo or our quarterly tech trends report for the firm. I’ll outline my thoughts, maybe bullet out some key insights, and then let Notion expand one of the sections. Think “assistant PM” that writes paragraphs when I’m too drained to wordsmith. If it sounds a bit robotic, I can refine it, but at least I’m not starting from zero. It’s also great for polishing: I can select a clunky sentence and prompt “make this sound confident” or “simplify this”, and it’ll suggest an alternative. It’s like having an editor who doesn’t mind my 2 AM writing habits.

At this point, you might imagine my desk like a scene from a hacker movie – multiple monitors, each with a different AI tool running, while I juggle coding, chatting, and analyzing. Honestly, some days it does feel like that! I’ll be merging a pull request with Copilot’s help on one screen, answering a partner’s question via a ChatGPT-crafted blurb on Slack in another, all while Devin builds an app in the background. It’s multitasking on caffeine and AI adrenaline. But remarkably, it works. These tools augment my workflow so I can context-switch without losing as much momentum.

Crucially, they free up my brain for the high-level stuff. Instead of spending all afternoon crunching a spreadsheet or combing through documentation, I can focus on making judgments: Does this tech align with our investment thesis? How do we help this founder solve their scaling problem? What’s the next strategic project that will give our firm an edge? AI doesn’t make those decisions for me, but it clears the brush so I can see the path. It’s like having a team of diligent (if slightly quirky) associates who handle the prep work, while I get to do the fun part of connecting dots and thinking big picture.

There’s also a fun side-effect: using these AI tools every day keeps me sharp and curious. I’m constantly experimenting, which means I’m always learning – whether it’s a new framework Copilot suggested or a new question about ethics because ChatGPT raised an interesting point. In an industry where "pattern-matching” is king, having exposure to so many AI-assisted insights widens my patterns. I spot red flags faster and validate gut feelings with more data. It’s like leveling up both as an engineer and as an investor at the same time.

Until next time!

Signing off and signing zero checks,

SWEdonym

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

Sign up and start mastering AI