toolmint

Author: ToolMinT

  • GPT-5.5 Just Dropped — Here’s What the Numbers Say Before I Test It Against Claude Code

    GPT-5.5 Just Dropped — Here’s What the Numbers Say Before I Test It Against Claude Code

    GPT-5.5 Just Dropped — Here’s What the Numbers Say Before I Test It Against Claude Code

    Content mode: Informed — Field Report

    I haven’t used this yet, but here’s what the data shows. OpenAI released GPT-5.5 yesterday (April 23, 2026), and as someone who already pays for both Claude Pro and ChatGPT Plus — about $40 a month to keep two AI writing tools open on my desktop — my first question wasn’t whether it’s “smarter.” It was: does this change which tool I leave open on which monitor?

    I’m not a developer. I write client deliverables — proposals, brand strategy docs, research briefs — and I use Cursor for the small scripts that clean up data between calls. I’ve been watching Claude Code from the outside the way most solo operators do: curious about where the coding-agent frontier lands, because when it lands well, it eventually trickles down to the tools non-devs like me actually use.

    So — what’s actually new in GPT-5.5, and where might it pressure Claude Code?

    What OpenAI actually shipped

    The release lands with concrete benchmark numbers, which is the first honest thing I’ll say about it: this isn’t a vibes announcement.

    • Terminal-Bench 2.0: 82.7% — the “operate the computer” benchmark, not just write code
    • SWE-Bench Pro: 58.6% — real software engineering tasks, the harder cousin of SWE-Bench
    • Codex context window: 400K tokens inside the coding environment; 1M tokens via the API
    • Token efficiency: OpenAI says GPT-5.5 “uses fewer tokens in Codex tasks” than GPT-5.4 at comparable quality
    • Latency: roughly matches GPT-5.4 per-token despite the capability gains

    Pricing for API access:

    Model Input / 1M tokens Output / 1M tokens
    gpt-5.5 $5 $30
    gpt-5.5-pro $30 $180

    Fast mode is about 1.5x faster at 2.5x the cost. Availability: rolling out to Plus, Pro, Business, and Enterprise in ChatGPT and Codex. GPT-5.5 Pro is Pro/Business/Enterprise only. API access “very soon.”

    The framing from OpenAI is explicit: this is a bet on agentic work. Greg Brockman called it “a real step forward towards the kind of computing we expect in the future.” Per TechCrunch’s reporting, OpenAI is openly talking about bundling ChatGPT, Codex, and an AI browser into a single “super app” for enterprise customers.

    Terminal running an AI coding agent on a laptop
    Terminal running an AI coding agent on a laptop

    Why the Claude Code comparison matters now

    Claude Code, as it stands while I’m writing this, is Anthropic’s agentic coding CLI that sits on Claude Opus. It’s what developers reach for when they want an AI to actually do work in the terminal — run commands, edit files, chain steps, return a diff. It’s had roughly a year to earn trust among people who care about that loop behaving predictably.

    GPT-5.5 in Codex is now the direct rival to that workflow, not just ChatGPT-the-chatbox. That’s the shift.

    Two things jump out before I’ve run a single side-by-side:

    First, 82.7% on Terminal-Bench 2.0 is the real claim. A coding chatbot can ace SWE-Bench in a controlled loop. Getting 82.7% on “use the computer like a person” is a different problem — navigating folders, reading error messages, recovering from the model’s own bad decisions. If that number holds up in the wild, it’s the first time I’d credibly hand a messy local task to an OpenAI agent and not feel like I was gambling.

    Second, 400K Codex context is enough for real work. The friction I’ve watched developer friends hit with agentic coding tools is rarely “the model is dumb.” It’s “I had to re-paste context twice because the window filled up.” 400K is generous enough that a mid-sized project fits without stunting the loop.

    What I’d want to know before reading further into the hype: how GPT-5.5’s agentic loop handles refusing to do things it isn’t sure about. Claude Code’s tuning leans cautious — it stops and asks when the plan could destroy your repo. The “super app” framing from OpenAI worries me slightly in the opposite direction. A confident, wrong agent is worse than a cautious, slow one.

    Where I predict GPT-5.5 will pressure Claude Code

    Trying to be specific instead of making a vibes call. Three places I expect GPT-5.5 to put heat on Claude Code in the next few months:

    1. One-off computer tasks. “Open this folder, rename 300 files based on a CSV, verify the output.” The Terminal-Bench 2.0 score suggests GPT-5.5 is built for exactly this. For me, that’s the weekly client-deliverable packaging job that currently eats 40 minutes in Cursor.
    2. Research workflows that blend code and writing. OpenAI is explicitly pitching scientific and technical research. “Write a Python script to scrape data, then draft the summary memo” is a natural fit when both halves live in one model. Claude Code is excellent at the code half; Claude.ai is excellent at the writing half; having both in one workflow has always required switching windows.
    3. Solo operators who already live in ChatGPT. If you pay for Plus and use Claude only occasionally, GPT-5.5 inside Codex removes a reason to keep two subscriptions. The gravitational pull toward consolidation is real when your monthly AI spend is already past $100.

    Where I predict Claude Code keeps the edge

    1. Long-form writing judgment. Speculation, but Claude’s emphasis on writing quality has held across four or five model generations. GPT-5.5 has to prove it’s closed that gap on long deliverables, not just matched benchmark scores.
    2. Cautious refusal behavior. When I don’t want an agent to take a wild swing at my files, Claude still feels like the safer default.
    3. The next Anthropic move. GPT-5.5 Pro at $30/$180 per million tokens is aggressive pricing for a frontier model. The real question is what Anthropic ships next — and whether yesterday’s announcement pressures them to move the Claude Code story forward faster.
    Solo operator workspace with dual monitors, testing AI tools side by side
    Solo operator workspace with dual monitors, testing AI tools side by side

    What I’m actually going to do this week

    I’m not switching defaults on client work. The cost of getting a deliverable wrong is higher than the cost of running two subscriptions for another month.

    What I will do:

    • Run the same 4,000-word brand strategy draft prompt through GPT-5.5 and my current Claude workflow. Compare coherence around sections 4-5, where ChatGPT has historically lost the thread for me.
    • Hand GPT-5.5 inside Codex a real file-rename-by-CSV task I’d normally do in Cursor. Measure wall-clock time and whether I had to correct anything.
    • Wait two weeks. First-week benchmarks always lead; first-month behavior is what actually matters.

    I’ll write that up with real numbers when I have them. Until then, treat this as what it is: a public-sources Field Report on a model I haven’t run.

    FAQ

    Is GPT-5.5 available in the ChatGPT Plus tier I already pay for?

    Yes. OpenAI confirmed rollout to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex. GPT-5.5 Pro (the higher-tier version) is Pro, Business, and Enterprise only.

    Does GPT-5.5 replace Claude Code for agentic coding?

    Not yet, and probably not for everyone. Claude Code’s strength isn’t raw benchmark — it’s the loop’s tuning and a year of field use. GPT-5.5’s 82.7% on Terminal-Bench 2.0 is the first time OpenAI has credibly entered that space head-on, but “best on paper” and “best to live with” aren’t the same thing.

    How does GPT-5.5 pricing compare for a solo operator?

    At $5 input / $30 output per million tokens, standard GPT-5.5 isn’t the cheapest frontier model on the market, but it isn’t the most expensive either. Unless you’re burning through millions of tokens a week, the subscription tier matters more than API pricing — and for most freelancers, that means ChatGPT Plus at $20/month.

    Should I cancel Claude Pro?

    Not based on benchmarks alone. I’m keeping both for at least another billing cycle. If you already use ChatGPT more, this release gives you cover to consolidate. If you use Claude primarily for writing, give it a month before moving anything.

    Sources

    AI-assisted research and drafting. Reviewed and published by ToolMint.

  • Cursor for Non-Developers: A Freelancer’s First Month With AI Coding

    Cursor for Non-Developers: A Freelancer’s First Month With AI Coding

    Laptop with code editor open, freelancer learning to automate small scripts

    I watched a colleague type a sentence into her laptop and watch 380 mis-named photo files rename themselves in under a minute. I’m not a developer. She isn’t either. I’d been planning to batch-rename the exact same kind of folder for a client that afternoon, and I’d budgeted three hours for it in Finder.

    That’s what made me subscribe to Cursor. I expected to cancel within two weeks. I’m still paying for it, and I want to tell you what actually happened, because most Cursor coverage assumes you already write code for a living.

    This is for the freelancer who keeps thinking “I bet I could automate that if I knew how to code.” You can, mostly. And you don’t need to learn to code the way you think.

    ## What Cursor actually is, in plain terms

    Cursor is a code editor — like opening a text file with formatting and folder navigation. The thing that makes it different from a normal editor is that it has a chat panel and an “agent” mode that can read your files, write new ones, and run commands on your computer. You can describe what you want in English, and it produces (usually working) code, edits files for you, and tells you what it did.

    For a non-developer, that means: you can sit in front of a folder of files, type “rename all these PDFs from `IMG_1234.pdf` to `[date] [original name].pdf` based on their creation dates,” and watch it happen. You don’t write the script. You don’t even need to read the script unless you want to.

    The skill you need is *describing what you want clearly* and *checking the result*. That’s it.

    [SCREENSHOT: Cursor’s chat panel on the right, showing a plain-English request and the agent’s plan to handle it, with a folder of test files visible on the left]

    ## My week-one setup (45 minutes total)

    I wasted an hour my first day trying to “configure” Cursor properly. Don’t. The defaults are fine for non-developer use. Here’s the setup that actually mattered:

    1. **Install Cursor** from the official site. Open it.
    2. **Sign in** to the AI features (free tier exists; I subscribed because I wanted the better model). Don’t bother with the free tier if you’re going to use this seriously — the pricing tier difference matters for the agent mode.
    3. **Make a folder somewhere** like `~/cursor-experiments` and open it in Cursor (File → Open Folder). This is your sandbox. Never run agent commands inside folders that contain files you can’t afford to lose, until you trust it.
    4. **Skip the extensions, themes, settings tabs.** You don’t need them yet.
    5. **Open the chat panel** (default keyboard shortcut shows it on the right) and try one tiny task: “Create a file called `hello.txt` with my name on the first line.” Watch what happens. Read the result. Confirm.

    That’s it. The first three days are about **building trust by giving it tiny tasks and verifying every output**. Treat it like a new freelancer you just hired — you’d check their first deliverables carefully before handing them anything bigger.

    Safety rule, non-negotiable: never run Cursor’s agent mode inside a folder that contains files you can’t afford to lose — until you’ve built trust with a week of sandbox work. It writes, renames, and deletes files. This is not ChatGPT.

    Content mode: Tested — I use this

    ## Three small wins that paid for the subscription

    **Win 1: Bulk-renaming a client’s photo deliverables.** A photographer client sent me 380 image files with iPhone-default names. I needed them named `[shoot-name]-[YYYY-MM-DD]-[001].jpg` and grouped into subfolders by date. I described that in the chat. Cursor wrote a Python script, asked me to confirm before running, and reorganized the entire folder in 12 seconds. That task would have taken me a half-day in Finder.

    **Win 2: Cleaning a messy CSV from a content audit.** A client gave me a 4,000-row spreadsheet of URLs, some malformed, some duplicated, some with weird whitespace, several columns I didn’t need. I described what “clean” should look like. Cursor produced a cleaned CSV plus a one-page summary of what it removed. I spot-checked 20 random rows. Three hours of work, done in five minutes.

    **Win 3: Scraping public pricing pages for a research deliverable.** I needed competitor pricing for a strategy doc — 14 SaaS companies, public pricing pages. I gave Cursor the URLs and asked for a tabular summary. It wrote a small script, ran it, and gave me a markdown table. About half the rows needed manual fixes (some pages had unusual layouts). Total: 25 minutes vs the 2 hours of copy-paste it would have been.

    Three wins, three to four hours saved, in the first week. The subscription paid for itself before the trial period was up.

    ## The traps that waste hours if you’re not careful

    **Trap 1: Treating it like ChatGPT with autocomplete.** Cursor’s agent mode actually does things on your computer. It runs commands. It writes and overwrites files. ChatGPT just talks. The mental model has to be “I’m directing a careful intern who has full keyboard access,” not “I’m chatting with an assistant.”

    I learned this when I asked it to “clean up this folder” without specifying what to keep. **It deleted three months of working notes.** They were in version control by luck. Always say what to keep, where to put output, and *test in a sandbox folder first*.

    **Trap 2: Accepting code you don’t understand the gist of.** You don’t need to read every line. But before running anything that touches more than one file, scan the agent’s plan and ask “is this doing what I expect, roughly?” If you can’t tell, ask it to explain in plain English first. It will. The five seconds of asking is the cheapest insurance you’ll ever buy.

    **Trap 3: Letting the project sprawl.** Cursor invites you to start “one little script” that becomes a half-built application you don’t understand. For non-developers, the right discipline is: every project starts in its own folder, has a one-line README explaining what it does, and gets deleted when it’s served its purpose. Don’t accumulate.

    **Trap 4: Skipping version control on anything you keep.** If a script becomes useful to you (you’ll run it again next month), put the folder in a free GitHub repo before you continue editing. Cursor will help you set this up if you ask. Without version control, an “improve this” prompt that goes wrong takes your prior working version with it.

    ## What still feels too hard for a non-developer

    I want to be honest about where I’ve hit walls:

    – **Setting up local development environments** for actual web apps. Even with Cursor’s help, getting a local Next.js or Rails environment running involved enough terminal pain that I gave up twice.
    – **Understanding error messages from runtime failures.** Cursor helps a lot, but some errors are genuinely cryptic and require either patience or asking a developer friend.
    – **Knowing what’s a reasonable scope.** “Build me a tool that does X” with a vague X leads to half-finished projects. Tight, specific, one-task scripts work great. Medium-sized “tools” do not.

    If you’re a freelancer hoping to build a SaaS product solo using Cursor — possible, but you’re learning a lot of unfamiliar things, not just delegating to AI. If you’re hoping to automate small parts of your existing workflow, you’re in the sweet spot.

    ## Who it’s for, who it isn’t

    Worth the subscription if any of these are true:
    – You regularly do file-shuffling tasks (renaming, sorting, converting formats)
    – You wrangle messy data from clients (CSVs, spreadsheets, exports)
    – You wish you could “just script that” but never learned how
    – You’re curious about coding and want a low-floor on-ramp

    Probably not worth it if:
    – Your work doesn’t involve repetitive file/data tasks at all
    – You’d rather pay a freelance developer to handle the occasional script
    – You don’t enjoy the small detective work of understanding what an automation is doing

    ## FAQ

    ### Is Cursor better than just using ChatGPT for code questions?

    For asking a question about code or getting a snippet to copy: ChatGPT or Claude is fine. For *running* the code on your computer, editing files in place, and watching the result without copy-paste — Cursor is in a different category. The difference is whether you want answers or actions.

    ### How much should I budget per month?

    The paid tier I’m on is in the $20-30/month range — check current pricing because they’ve been adjusting it. For a freelancer using it 3-5 times a week, it pays for itself in time saved. For occasional use, the free tier or paying per use might be better.

    ### What if it deletes something important?

    Mistakes will happen. Defensive setup: keep a sandbox folder for experiments, use version control (git, even just locally) on anything you’d hate to lose, and make sure your computer’s regular backup (Time Machine, Backblaze, whatever) actually runs. Treat agent-mode AI like a power tool — useful, dangerous near your data, worth using with both hands and a clear workspace.

    —
    *AI-assisted research and drafting. Reviewed and published by ToolMint..*

    Cursor Pricing for Freelancers

    Plan Price AI Usage
    Hobby Free Limited agent requests and tab completions
    Pro $20/month Extended agent limits, frontier models, cloud agents
    Pro+ $60/month 3x usage on OpenAI, Claude, Gemini models
    Ultra $200/month 20x usage, priority access to new features
    Teams $40/user/month Shared chats, centralized billing, SAML SSO

    My recommendation for non-developers: Start with the free Hobby plan for 2 weeks. If you’re hitting the agent request limits regularly, the Pro plan at $20/month is worth it. Pro+ and Ultra are overkill unless you’re running AI coding tasks all day.

    Try Cursor free →

  • Perplexity vs Google for Research: How I Cut Pitch Prep from 90 Minutes to 15

    Perplexity vs Google for Research: How I Cut Pitch Prep from 90 Minutes to 15

    Researcher comparing search results across browser tabs on a laptop

    I used to count my pitch prep in browser tabs. Twenty-two felt normal. Company site, three competitor sites, a couple of industry reports, two or three LinkedIn pages, the founder’s podcast episode I half-remembered, a press release someone mentioned in a Slack DM. Most of those tabs were open for 30 seconds and closed without me learning anything — I was padding effort hoping something useful would jump out.

    That stopped when I started using Perplexity. Not because it replaces research. Because it compresses the *first pass* — the tab safari — from 90 minutes into a structured 15-minute Q&A with citations I can verify.

    This is the workflow I’ve settled into after about a year of running the two side by side on real proposals.

    ## The 15-minute pre-pitch pass

    Before any first call with a prospect, I run the same five questions in Perplexity. Each takes about 90 seconds to produce a cited answer.

    1. “What does [company name] do, in one paragraph, based on their public statements in the last 12 months?”
    2. “Who are [company]’s named direct competitors and how do they position differently?”
    3. “What’s been in the news about [company] in the past 6 months?”
    4. “What do customer reviews on G2, Capterra, or Reddit say about [company] (positives and complaints)?”
    5. “Has [company] published or said anything about [my service area] recently?”

    Each answer comes with citations I can click and verify. I copy the citations into my call prep doc so I have receipts if the client asks “where did you see that?” Total time: usually under 12 minutes.

    Why this works: the first pass isn’t meant to be comprehensive. It’s meant to surface where the second pass should spend its time. Skipping this step is why your pitch prep takes 90 minutes.

    Content mode: Tested — I use this

    The output isn’t comprehensive — that’s the point. It’s a *first pass* that tells me where to dig deeper for the next 20 minutes if needed.

    [SCREENSHOT: Side-by-side of a Perplexity answer with citation chips and the same query in Google showing 10 ads + 3 organic results]

    ## Where Perplexity actually wins

    The wins are specific:

    **Source citation built in.** Every claim has a numbered citation. I can click straight to the original page. With Google, I read 5 pages and try to remember which one said what.

    **Synthesis across sources.** When I ask about a company’s positioning, Perplexity often pulls from their site, an interview the founder gave, and a press release — and synthesizes the three into one coherent paragraph. With Google, I read three pages and synthesize in my head. Faster, less error-prone in Perplexity for first-pass.

    **Fewer ads in my way.** Google’s commercial intent detection is aggressive. Searching “[company] reviews” gives me 4 ads, sponsored placements, then organic. Perplexity has no ads. The signal is denser per minute.

    **Q&A as a default mode.** I think in questions during prep, not keywords. “What problems does this company seem to be solving for their customers based on their case studies?” is a sentence I can type into Perplexity verbatim. Translating that into Google keywords loses information.

    ## Where Google still wins

    Three categories where I switch back without thinking:

    **Recency and breaking news.** If something happened in the last 48 hours — a layoff, a funding announcement, a product launch — Google News still surfaces it faster and more completely. Perplexity is improving but lags on freshness for time-sensitive prep.

    **Local intent.** “Best [profession] in [city]” or anything where I want a *map result*, opening hours, or a directions link. Perplexity isn’t built for this and shouldn’t be.

    **Vendor-specific deep dives.** When I need to know exactly what a SaaS product’s pricing tier includes, going to their pricing page directly via Google search is faster than asking Perplexity to summarize it (and getting an answer that might be 6 months stale).

    **Image search.** Visual research, mood references, finding a specific photo someone posted somewhere — all still Google or specialized image tools.

    The pattern: Perplexity for understanding, Google for finding.

    ## My repeatable proposal research template

    This is the doc I duplicate for every new prospect. It lives in my Notion as a template page.

    “`
    ## Prospect: [Company Name]
    Date: [today]
    Source of lead: [referral / inbound / cold]

    ### Quick read (Perplexity, 12 min)
    1. What they do (1 paragraph + citations):
    2. Competitors (3-5):
    3. Recent news (last 6 months):
    4. Customer sentiment:
    5. Published views on [my service]:

    ### Deep dive (manual, 20 min, only if call is confirmed)
    – Their team page — who I’d likely work with
    – Their case studies — proof points I can reference
    – Their pricing or positioning if visible — sets the conversation tone
    – LinkedIn check on the call attendees — recent posts only

    ### Hooks for the call
    – One specific thing they’re doing well
    – One specific thing they might be struggling with (positioned as a question, not a critique)
    – One question only I would think to ask
    “`

    The Quick read is Perplexity. The Deep dive is browser tabs. The split is what saves the time.

    [SCREENSHOT: The template doc above, with a real (anonymized) prospect’s research filled in]

    ## Two traps I had to learn the hard way

    Heads up — citations look authoritative but aren’t proof. Read the next two paragraphs before you paste anything from Perplexity into a client doc.

    **Citation laundering.** Perplexity citations look authoritative because they’re presented in clean little numbered chips. They’re not always right. I’ve had Perplexity cite a Reddit thread for a “fact” that was speculation in the thread. **Always click through citations** on anything you’ll repeat to a client.

    **Stale answers.** Perplexity caches some answers and serves them when the underlying source has updated. If recency matters, append “as of [current month/year]” to the question or open the original source.

    A two-minute habit fixes both: don’t paste anything from Perplexity into a client doc without clicking at least one citation per claim.

    ## Two real prep sessions, side by side

    To make this concrete, here are two recent prep sessions where I tracked time:

    **Session A — boutique branding agency, 35 employees, B2B services prospect.** Lead came in by referral. I had 24 hours before the call. Old workflow: I’d have spent ~75 minutes across the agency’s site, three case studies, two LinkedIn pages, and a podcast they’d recently been on. New workflow: 11 minutes in Perplexity for the five-question pass, then 18 minutes manual deep-dive on the two case studies most relevant to the kind of work I do. Total: 29 minutes. The call went well. I closed the deal a week later, partly because I name-dropped a specific case study insight in the first 10 minutes.

    **Session B — early-stage SaaS, founder-led, cold inbound.** Less to research because the company is younger. Old workflow would have been 30-40 minutes regardless — I’d have padded with industry-trend reading. New workflow: 7 minutes in Perplexity got me everything public, plus a flag that the founder had recently published a critical post about a competitor. I read that post (5 minutes) and showed up to the call already aware of his stated views. Total prep: 12 minutes. The conversation was sharper and shorter, which the founder explicitly thanked me for.

    The pattern across both: Perplexity didn’t replace any *necessary* deep work. It killed the unnecessary skim time so I had budget for the depth that mattered.

    ## The time math, honestly

    If you regularly research prospects, competitors, or client industries — and the research is currently a tab-explosion problem — Perplexity Pro is one of the highest-leverage tools per dollar in my stack. Total time saved per proposal: about 90 minutes, conservatively. Across 4-6 proposals a month, that’s a billable day reclaimed.

    If your research is mostly local, mostly visual, or mostly chasing breaking news, Google still does the job and you don’t need a second tool.

    ## FAQ

    ### Do I need Perplexity Pro or is the free tier enough?

    Free is fine for trying the workflow. The Pro tier matters if you want longer/deeper answers, file uploads (handing it a PDF prospectus and asking questions), and access to better models. For occasional pitch prep, free works. If you do this weekly, Pro pays for itself in saved time.

    ### What about ChatGPT’s web search or Claude’s web search?

    Both work. The differences as of when I’m writing this: Perplexity is the most “search-first” interface — citations are first-class and the UI is built around the cite-and-verify loop. ChatGPT/Claude search is more conversational; you can refine in dialogue but citations are less prominent. For quick research I prefer Perplexity. For long analysis where I’m reasoning through findings, the conversational tools are better.

    ### Can I trust Perplexity’s answers for client-facing claims?

    Trust the citations, not the synthesis. If a Perplexity answer says “Company X has 200 employees” with a citation, click the citation. If the source confirms, you can repeat it. If you can’t find the claim in any cited source, don’t repeat it. This is the same hygiene a careful researcher applies to any secondary source.

    —
    *AI-assisted research and drafting. Reviewed and published by ToolMint..*

    Perplexity vs Google: What Each Costs

    Tool Free Paid
    Google Search Free (unlimited) Google One AI Premium: $19.99/month
    Perplexity 5 Pro searches/day $20/month — unlimited Pro searches, file uploads

    Worth paying for Perplexity Pro? For freelancers doing client research more than three times a week, yes. The ability to attach documents, run Pro searches, and get cited sources cuts research time significantly. I’ve had it pay for itself in a single pitch prep session.

    Try Perplexity Pro →

toolmint
Powered by  GDPR Cookie Compliance
Privacy Overview

This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.