Your year in AI conversations
748,607 total words exchanged
Note: Jan-Aug shows split usage between Claude and ChatGPT. September was almost exclusively ChatGPT. Oct-Dec reflects full-time Claude usage.
Claude.ai web usage only — API and Claude Code usage would add to this total
Your 2,100 message exchanges produced an estimated 87.3 kg of CO2. That's 36.9 kg from inference + hardware embodied carbon, plus 50.4 kg from amortized model training. Equivalent to driving 218 miles. Data center and power plant cooling used 546 liters of water (8.4 showers).
To neutralize your Claude.ai carbon footprint, donate $1.75 to a quality offset provider: Gold Standard, South Pole, or Cool Effect.
Aggressive methodology for power users: 20 Wh/query (above 17 Wh extended thinking benchmark), ×1.3 PUE, ×1.5 hardware embodied, 0.45 kg CO2/kWh grid, training at 24g/query (6g base × 4 for reasoning overhead), water at 10 L/kWh (direct + indirect), $20/ton offset. Sources: arxiv.org/abs/2505.09598, EESI, IEEE Spectrum
You use Claude less for research and more for processing. Working through pricing decisions, drafting emails, pressure-testing strategies. Your pattern: present something, get a reaction, refine. It's externalized thinking - you're clearer on what you want after you've seen a response to it.
80% of your messages are under 50 words. You don't over-explain upfront - you share context, get a response, then steer with "tweak this" or "not quite." 44 conversations went past 20 messages this year. You're willing to iterate when it matters, but you don't waste words getting started.
Most of your conversations aren't questions - they're working sessions. You draft proposals, emails, and strategies in real-time rather than planning in isolation. You seem to think better when you have something concrete to react to and refine. The output is the thinking process.
Only about 12% of your conversations are pure information lookups. The rest are working sessions - drafting, strategizing, decision-making. You use Claude more like a colleague you can think out loud with than a reference tool.
You share work-in-progress early, get feedback, incorporate it, repeat. It's efficient - you don't spend time explaining context that becomes obvious from the draft itself. You're not precious about first attempts; iteration is the point.
The year started with Solstice Aerospace front and center - 79 aviation-related mentions in February. Pitch deck work, cofounder search, hydrogen aircraft feasibility research. You were heads-down on one thing.
55 conversations - your busiest early month. Aviation work continued (recruiting engineers, accident analysis, Pacific logistics), but life stuff appeared too: a child's stomach symptoms needed sorting out. The dual-track of work and parenting showed up clearly this month.
Aviation still central, but other things started appearing. First vermouth conversations. A traffic light alert device concept. Lost jewelry, insurance claims. Less singular focus, more variety creeping in.
Focus shifted to positioning yourself: superpower statements, bio drafts, website CSS. A lot of Ghost theme work. You were building the infrastructure for fractional consulting - how you'd present yourself to potential clients.
Only 24 conversations across two months - noticeably slower. You were splitting time more aggressively between Claude and ChatGPT during this period. The conversations that happened were about Duckbill marketing, consulting landing pages, new client work. A pivot was underway, just not a loud one.
1 conversation. You switched to ChatGPT almost exclusively for the month - an experiment in the other direction. The lone Claude conversation that month was brief. This is the gap in the chart.
119 conversations - you came back from the ChatGPT experiment and went all-in on Claude. Duckbill work dominated (74 mentions), but you were also into fitness tracking, food logging apps, parenting stuff, vermouth experiments. A lot of plates spinning simultaneously.
140 conversations - your highest month. Vermouth/Cartographer became a real focus (76 mentions). Conversations with the Four Sigmatic founder. Restaurant recommendations. Parenting. Style advice. Kid was sick. You were doing a lot of different things at once.
54 conversations - slower pace. Technical projects, partnership thinking, this Wrapped analysis. More reflection than execution. Setting up for next year rather than sprinting to finish this one.
You started 2025 focused on one venture (Solstice). By fall, you were running multiple tracks: Duckbill consulting, vermouth development, fitness optimization, parenting, personal projects. Whether this was intentional strategy or natural drift, it's where you ended up.
Your October-November activity spike (259 conversations in two months) wasn't followed by burnout - December was productive, just calmer. You seem to handle variety without it becoming chaos. The open question for 2026: is the multi-track approach sustainable long-term, or does something eventually need to become the main thing?