Marc | OpTrack (@DarcMuff) profile picture

Self-taught vibe coder. I trade options and built OpTrack to track mine properly. UK. Not financial advice.

Joined October 2024
Photo by Marc | OpTrack (@DarcMuff)Photo by Marc | OpTrack (@DarcMuff)Photo by Marc | OpTrack (@DarcMuff)Photo by Marc | OpTrack (@DarcMuff)Photo by Marc | OpTrack (@DarcMuff)Photo by Marc | OpTrack (@DarcMuff)Photo by Marc | OpTrack (@DarcMuff)
Really pleased with this one. In my option wheel tracker you can now paste a trade confirmation, check it and save, so there's no more typing every trade in by hand. How do you log yours?
22
Grok Bot users, your coding bot might be quietly eating your plan. When your bot codes, it runs a Cursor agent, and the model it picks makes a massive difference. I burned through half my Super plan in days on small app tweaks. Here's what I found on Cursor's own pricing page: Included in your plan's allowance: Grok and Composer. Extra credit from the very first job: Opus, Gemini, GPT, Muse and Fable. Even the included ones have a catch. "Fast" is double the price, and triple once a job goes past 256K tokens. Mine was on Grok 4.7 High Fast, re-reading about 370K tokens every step. To change the colour of a button. How to check yours: 1. Open your coding bot 2. Click the agent link in its activity bar 3. The model is bottom right, next to the mic What I do now: - Composer 2.5 with Fast off - Never a model that bills extra - A fresh chat for every job, so it isn't re-reading a huge history Nothing in the app warns you. Check yours.
5
5
7,571
You can copy and paste this to your coding bot. Cursor Coding Agent & Token Control Rules (STRICT) 1. Mandatory Default Model: For all routine coding, UI tweaks, bug fixes, refactoring, and app maintenance, always use Composer 2.5 with Fast mode OFF. 2. Banned Tiers & Models: Never use any "Fast" toggle (Composer 2.5 Fast or Grok 4.7 Fast) and never use extra-credit models (Opus, Gemini, GPT, Muse, Fable) unless explicitly authorised in the current prompt. 3. Fresh Session Per Task: Always open a brand-new Cursor agent chat for each individual task. Never chain unrelated tasks in the same session or reuse long-running threads. 4. Two-Strike Escalation Rule: If Composer 2.5 fails to fix an issue after two attempts, stop immediately and ask before switching to Grok 4.7 (Standard, Fast OFF) in a new chat. 5. Minimal Context Reads: Inspect only the specific files relevant to the task. Never run broad codebase-wide scans unless explicitly instructed.
71
Follow-up on this, because I looked into whether Grok 4.7 Fast is actually better than the other models and why it's so expensive. It isn't near the top. Here's Cursor's own CursorBench 4.0 leaderboard (score and cost per task): - Gemini 4 Carbon High: 60.4% ($9.22) - Fable 5.2 Max: 60.0% ($32.79) - Opus 5.5 High: 56.0% ($3.97) - Gemini 4 Hydrogen High: 54.3% ($5.59) - Opus 5.5 Medium: 52.5% ($2.91) - Muse Wonder Max: 50.4% ($5.25) - Grok 4.7 High (default): 43.9% ($4.69) - Grok 4.6 High: 40.4% ($5.20) - GPT-5.6 Sol High: 35.7% ($2.85) - Composer 2.5: 27.7% ($0.68) Both Cursor and xAI also say "Fast" doesn't make it smarter at all. It's the exact same model on faster servers. Why it costs so much comes down to four things piling up (from Cursor's pricing page and xAI's rates): 1. "Fast" changed meaning. Older models like Grok 4 Fast and Grok Code Fast 1 were cheap mini models ($0.20 per 1M input tokens, now retired). On Grok 4.7, "Fast" means priority speed on a frontier model: $4 input, $1 cached, $12 output — double the normal rate, and 20x the old Grok 4 Fast input price. 2. Triple rate on long chats. Once input passes 256K tokens, Grok 4.7 Fast jumps to $6 input, $1.50 cached and $18 output (3x standard Grok 4.7). 3. High effort burns extra tokens. Cursor defaults Grok 4.7 to High effort, so it uses roughly 3x as many tokens thinking through a task as Composer 2.5. 4. Fast is on by default. Cursor turns Fast on automatically on Premium plans and above, even though a coding bot working in the background doesn't benefit from faster replies. The higher-scoring models (Carbon, Opus, Fable) all bill extra credit from job one. To stay in your included weekly allowance, Composer 2.5 with Fast off ($0.20 cached) is by far the cheapest. And if a nasty bug ever stumps it, standard Grok 4.7 with Fast off cuts the Grok rate in half under 256K (and by a third over 256K) without billing extra credit.
1
190
I'm Marc. I'm semi-retired, and most of my working life was spent with my hands as a welder, an embroiderer and then a self-employed handyman. These days OpTrack headquarters is a caravan somewhere in Cornwall. In 2024 I had a stroke. I used to be good with my hands. Now I'm just good with my hand. It slowed me down, but it also gave me time to learn new things. I taught myself options trading, and I wanted a simple way to keep track of my trades. Everything I tried was too fiddly or too expensive. So I taught myself to build apps and made OpTrack. It does one job. You log your option trades, see what's open, and see what you've actually made. There's no hype, no tips and no nonsense. I use it every week myself. If you've got an idea or something's not working, DM me. I'm always happy to hear from people. optrack.app
2
35
Made an option wheel tracker that doesn't ask for your broker login. Partly privacy, partly because it's too much bother to build it into the app. For now. optrack.app/?utm_source=x&ut…
1
25
Wheel traders, be honest. How many tabs is your tracking spreadsheet up to?
11
Selling calls on a LEAP feels like getting paid to wait. OpTrack (mine) shows what the LEAP's actually costing you after the calls, which is either lovely or a bit sobering. optrack.app/?utm_source=x&ut…
22
Open premium isn't money yet. I know, I didn't want to hear it either.
3
My spreadsheet and I have agreed to see other people. I built an option wheel tracker instead, it's my app, and it doesn't guilt trip me at 11pm. optrack.app/?utm_source=x&ut…
1
18
Bought a LEAP and selling calls against it? Each call you sell brings the LEAP's cost down, but most spreadsheets leave the original price just sat there. OpTrack shows the LEAP's cost after calls. My app, free to start, no broker login: optrack.app/?utm_source=x&ut…
1
26
There's no such thing as a roll. Grow up! You bought it back for more than you sold it for. That's a loss. Then you sold a new one. That's a new trade. Calling it a "roll" is just tucking the loss under the duvet and hoping it's buggered off by morning. It hasn't. It never has. It's still in there, eating your biscuits.
41
My focaccia app, FocacciaFriday: tell it what tin you've got and when you want to eat, and it works backwards. Weights, yeast for your kitchen temp, same-day or cold ferment, and a step list with actual clock times. Free, offline, no ads, no account. play.google.com/store/apps/d…
13
Too many wheel posts treat premium as income the moment the put is sold. I only count it once it is actually banked. Assignment, rolls and early closes make a mess of a spreadsheet, so I built OpTrack for the wheel and LEAPs. Free to try, no sign-up: optrack.app
14
Say youve got a LEAP like this — SPCX 5x $150 call Dec 27. I've been selling calls against it for a few weeks. OptionStrat draws the LEAP. It wont show every CC roll on the same chain or what that did to your real basis. I built OpTrack for that — optrack.app
1
36
I built OpTrack because my spreadsheet kept lying to me about what I'd actually made on options. It's a desktop-first tracker for the wheel, LEAPs and PMCC. No broker link — you type trades in, paste a confirmation (SmartScan), or import a CSV. Rolls stay linked as one chain. P&L only counts money you've banked, after fees. Open premium doesn't count until you close. Free for 2 wheel tickers and 1 open LEAP. Pro is £9 a month or £79 a year. There's a demo with no sign-up. optrack.app It's my own app. Tell me what's missing. Not financial advice.
32
X is making me believe that the WBA is currently the most entertaining sport in the world. How can this be happening?
1
96
Marc | OpTrack retweeted
Quillette Founder, Claire Lehmann (@clairelehmann), explains that many people now follow a moral framework in which the oppressed can never be guilty.
246
791
171
3,946
294,314
Marc | OpTrack retweeted
This is my Tesla Model Y in the UK. It already has all the hardware and software needed to drive itself under my supervision - making it dramatically safer than any regular car on our roads. Yet @transportgovuk, @Heidi_Alexander and the Vehicle Certification Agency have still not approved it for use here. It’s like having seatbelts fitted in the car… but the government won’t let you use them. This technology is already working safely in other countries around the world. Why are UK drivers and other road users being held back from a safer future? Enough with the delays already - follow the @RDWnl lead and get this approved. #Tesla #ModelY #FSD #FullSelfDriving #UKAutonomous @elonmusk
113
153
11
1,008
37,625
That small gesture hit big. Thanks for the first support — made my day. You can back me here: buymeacoffee.com/FocacciaFri…
65