Claude Sydney Meetup Recap — Skills, Tips, and Quota Maxxing

On March 31, the 02Ship community gathered at Haymarket HQ (Level 2, 63 Dixon St, Haymarket) for our second Claude Sydney meetup. Over 50 builders, students, and AI-curious folks showed up for an evening of lightning talks, demos, and networking — all centered on getting more out of Claude.

Attendees at the Claude Sydney meetup March 31

Thanks to Haymarket HQ for hosting and providing the space.

Talk 1: How I Use Claude Code for Non-Coding Work

Bob Jiang (02Ship) kicked things off with a 30-minute talk on treating Claude Code as a capable junior team member — not just for writing code, but for any text-based work.

The core mindset shift: stop thinking of Claude Code as "autocomplete on steroids" and start thinking of it as a junior colleague who can read your project files, follow multi-step instructions, and iterate based on feedback.

Bob walked through three use cases he runs entirely through Claude Code:

Making slides — The presentation itself was built with Claude Code using a custom /create-slide skill. One command generates a full HTML slide deck with 02Ship branding, speaker notes, and timing.
Writing blog posts — Claude Code reads existing .mdx files in the project, matches the frontmatter format, and writes new posts that fit the established patterns.
Recording course videos — A full pipeline from script outline to final MP4: HTML slides get screenshotted with Playwright, narration is generated via Google TTS, and ffmpeg composes the final video. The /generate-course skill orchestrates the entire flow from a single PDF input.

The key takeaway: Skills — reusable Markdown files that live in .claude/skills/ — are what turn one-off prompts into repeatable workflows. If you do something more than three times and output quality matters, write a skill for it.

Bob's rule of thumb for when to create a skill:

You do it more than 3 times
Output quality must be consistent
It references specific files or templates
Others on your team will do it too

Talk 2: Claude Tips and Tricks

Joseph Hilsberg — lecturer at UNSW (Cryptocurrencies & Decentralized Finance), research assistant in digital identity, and self-described "AI Enjoyooor" — shared a rapid-fire set of practical tips organized around four themes.

Skills and Prompts

Joseph highlighted the growing skills ecosystem — reusable capabilities you can install into any Claude Code project with a single command (npx skills add <REPO_URL> --skill <SKILL_NAME>). Resources like skills.sh and ethskills.com are curating community-built skills.

For one-shotting big features or new projects, Joseph shared his workflow:

[Claude] Interview — let Claude ask you questions to clarify requirements
[Claude] Spec — Claude writes the specification
[Codex] Review — get a second opinion from another model
Code — then build

The key insight: instead of prompting Claude, let Claude prompt you. His go-to review prompt: "Review and grill me with any questions or uncertainties while we formulate a final version of the implementation plan."

Workflows

Joseph covered three workflow categories:

Software dev — the interview-spec-review-code pipeline described above
Research — Claude excels at synthesizing research. Feed it sources, have it extract claims and citations, attack hidden assumptions, then synthesize across sources. The goal: conclusions are earned, not guessed. He also recommended Auto-Claude-Code-Research-in-Sleep for overnight research sessions.
Agents — AI agents that don't just chat but act. Popular frameworks include Claude Code, OpenClaw, and Hermes. Orchestrators like Paperclip AI and Factory AI are pushing toward autonomous software development.

Tools — The Current Setup

Joseph shared his multi-model toolkit (as of March 31):

Tool	Cost	Best For
Gemini AI Pro	$32.99/mo	Multimodal (images, audio), 3D modelling via OpenSCAD
Claude Pro	$29/mo	Green-field spec building, plan mode, skills + tool calling
ChatGPT Plus	$30/mo	Good memory, economical with GPT 5.4 tiers, feature-by-feature building
MiniMax Std	$10/mo	1,500 requests/5hrs, economical agent model
Grok	Free	X/Twitter search and trend analysis
Perplexity	API-based	Online research
Qwen3.5-27b	Free (local)	Decent reasoning on 16-24GB M-series Macs

His take on Claude specifically: "The ultimate generalist. Best at tool calling and agentic skills." He noted the Claude Max tiers ($169/mo, $340/mo) for heavy users who need sustained Opus access.

Quota Maxxing Tips

The crowd favorite section — four tips for getting the most out of your Claude Pro subscription:

Plan around 5-hour resets — Structure your day so deep Claude sessions align with quota resets. Joseph's move: message Claude "hi" 3-4 hours before a planned lock-in session to trigger the reset timer.
Opus for Plan, Sonnet for dev — Use /model opusplan to automatically switch to Opus 4.6 in plan mode and Sonnet 4.6 for coding. Opus is deeper but slower; Sonnet is 2x faster and cheaper.
Handovers for extended sessions — When quota is running low mid-session, ask Claude to "Produce me a handover I can provide Codex with to re-commence work whilst your quota resets." Switch tools without losing context.
Token maxx — When your session quota is about to reset, burn the remaining tokens on automated workflows rather than letting them expire.

Top Tip

Joseph's number one piece of advice: "Just talk to it." Referencing Peter Steinberger's post "Just Talk To It — the no-bs Way of Agentic Engineering", the message is simple: stop overthinking prompts. Just describe what you want in plain language.

His closing slide — a collage of projects he's built with AI tools — drove the point home: "You can just do things."

Community Shout-Outs

During the open mic:

Dave Katague shared his personal AI toolkit and how he uses different models across his workflow. Dave also wrote up a detailed summary of the entire meetup — check it out on his Notion page.
Burin took the mic to share with the group as well.

Open Mic and Networking

After the talks, attendees shared their own Claude workflows and projects during the open mic session, followed by an hour of networking. The range of backgrounds — from crypto researchers to first-time builders — made for great cross-pollination of ideas.

Join the Next Meetup

We're back on Tuesday, April 21 at Haymarket HQ (same venue, 6:00 - 8:30 PM). This time the format is lightning talks — quick, engaging presentations covering AI SDK, building intelligent applications, and real-world AI engineering.

We're looking for speakers and sponsors — if you've got something to share, apply via the registration page.

It's free. Spots are limited.

Register for the April 21 meetup on Luma

See you there.