auDA .au Domain Rule Changes Affect Small Businesses

auDA wants to change the rules on .com.au domains. Small business will pay for it. Most Australian businesses got their .com.au or .net.au domain with nothing more than an ABN or ACN. That has been the rule for years and it works. auDA is now proposing to scrap that pathway. Under Recommendation 2 of its Licensing Rules Review, you would need a registered business name that exactly matches each domain you hold. No match, no domain. That is not a tidy-up. It is a rewrite of who is allowed to own an Australian web address. The registrar behind this petition estimates around 2.7 million of the 3.4 million .au registrations could be affected, with hundreds of millions in one-off and ongoing compliance costs, and a real risk of domains being deleted because a paperwork detail did not line up. If you run a business on a .com.au, this is your problem too. Sign the petition and tell auDA to keep section (f) of the current rules: https://c.org/ZfFNK6ynFL

With auDA’s new policy rules set to cost Australian businesses millions of dollars annually, this disastrous policy needs to be stopped. Last year, I purchased a domain name that placed the transaction well within the Top 10 highest domain name purchases ever recorded in Australia. (More on that in a future post buyersagents.com.au) I made this investment because the right domain name provides a significant benefit and contribution to a business’s long-term success. It represents brand equity, customer trust and market authority. I’m also extremely proud of the ecosystem we have built Dotto to help businesses secure the perfect domain for them. These proposed auDA changes not only undermine the security of these digital assets, but they also threaten to strip away the "close and substantial connection" rule that has protected Australian businesses and consumers for years.If this policy goes through, it will trigger an unprecedented wave of administrative friction, corporate vulnerability and with out doubt legal disputes over core business names. We cannot let a bureaucratic shift erase the value and security of Australian digital identities. Please join me in signing the petition to demand that auDA halts this policy change immediately. https://lnkd.in/dQnVDh8t

Our Claude system has our company credit card. It can go and buy leads on its own (with a 100 euro limit ofc 😅) Lead sourcing inside the Kinetyca Console runs end to end. This is what a single run looks like: 1 - Pick the source 👉 buy fresh leads, or pull from the bench: leads we already sourced for that client, sitting in Supabase and never shared across clients 2 - Pick the targeting 👉 set the ICP by hand, or reuse the ICP from the campaign that is converting right now 3 - Pick the providers 👉 decide which data providers this run is allowed to buy from 4 - Set the budget and the brief 👉 how many leads, what the run is allowed to spend, and free text for everything the filters cannot say 5 - The executor takes over 👉 a skill in the back end calls the provider APIs. The keys resolve from a vault at runtime, so Claude never holds a password 6 - It asks before it assumes 👉 anything unclear comes back as a question first. Then it buys, and the finished list lands in the campaign Two guardrails sit on top of every run: 1. The spending cap 👉 the budget the run cannot exceed, set before anything happens 2. The confirmation threshold 👉 above 100 euro, the console asks a second time before money moves When people hesitate about autonomous systems, it is almost always about money, and the usual answer is a promise to be careful. We would rather have guardrails you can point at. Caps, thresholds and double confirmations are how software has handled money for decades. Autonomy is only as wide as the guardrails around it. These are two of ours. Where would you set the threshold in your own system?

Static dashboards are becoming a liability, and the reason is expectations. Your users now talk to software. They ask a question in plain language and expect an answer. When they open your product and find a grid of charts they have to interpret themselves, it does not read as thorough. It reads as dated. The bar moved from show me what happened to tell me what is happening, what changed, and what I should do next — inside the product, not in a separate tool. For anyone shipping software right now, this is a roadmap question, not an analytics question. The analytics surface in your app is increasingly the surface your users judge your intelligence on. If a customer asked your product a question today, would it answer?

Build something agents can use. Then design something humans can own. Your agent just deleted a £2M client's account. Not because it was malicious; because it read "delete my account" when the customer meant "pause my subscription." It had CRM write access. The escalate queue had no owner. The ticket sat for six days. Vendors sell task-completion. Nobody ships the operating model: who defines done, who holds the keys, who owns the second check. A sandbox is one fence. One fence is single-point failure dressed as defence. Before go-live, answer four questions in writing: - Done: who defines it, by name - Keys: who grants, who revokes - Triage: who owns escalate, monitor and review - Second check: who outside the deploying team can say no Full essay: https://lnkd.in/egQKsGjS Building on recent thinking from Aaron Levie, Noam Brown, Dwarkesh Patel, Anca Dragan and Rohin Shah.

I agreed to write a landing page. Three weeks later, I was filing a notification with a federal data protection regulator. I have a web design diploma. Four years after finishing it, I emailed my old mentors to tell them they’d left out the entire back-end of running a business. ✏️ Here is what that ‘simple’ page actually turned out to be: A payment gateway—someone has to be contractually responsible for it. A public offer, because taking money without one is a decision rather than an oversight. A privacy policy that matches what the form really collects, not what a template assumed it would. A regulator notification. A CRM, so requests actually land where a human will see them. Automated handling, so the reply doesn’t depend on whether the owner happens to be awake. None of that is visible on the page. All of it decides whether the page works. ✏️ A shorter list decides it too: Where does the request go? Who gets notified, and how quickly? What does the buyer see in the sixty seconds after paying? Because that’s the window in which people quietly decide whether they’ve just been robbed. And the question nobody puts in a brief: can the owner run all of this in three months without you? That last one is the actual deliverable. A page that only works while its author is still answering messages isn’t a page—it’s a dependency with a nice hero image. The client thinks they’re buying a button. They’re buying everything that happens between a stranger and their business after the button. If you launched something recently: which of these did you find out about afterwards?

𝗠𝘆 𝗔𝗴𝗲𝗻𝘁 𝗥𝗲𝘁𝘂𝗿𝗻𝗲𝗱 𝗦𝘂𝗰𝗰𝗲𝘀𝘀. 𝗧𝗵𝗲 𝗕𝗿𝗼𝘄𝘀𝗲𝗿 𝗦𝘁𝗮𝘁𝗲 𝗦𝗮𝗶𝗱 𝗢𝘁𝗵𝗲𝗿𝘄𝗶𝘀𝗲 An agent clicks a checkout button. The browser tool reports success. The run completes without errors. But the page stays on the cart. Which result do you trust? This is a dangerous state in agent development. The action succeeded, but the outcome failed. A browser driver might accept a click that fails to navigate. A queue might accept a message that a consumer later rejects. If your test only checks the tool return value, you are testing the transport instead of the task. You must distinguish between three things: - The tool function returned normally. - The agent run completed normally. - The expected state transition occurred. The first two are execution evidence. The third is outcome evidence. You cannot swap them. To fix this, perform the action and then observe the world through a separate surface. Do not rely on the same tool that performed the work. Use these better observation methods: • Browser click: Check the DOM, URL, or accessibility tree. • Database write: Perform an independent read using a business key. • Queue publish: Verify consumer receipt or downstream state. • File generation: Check file existence and the content digest. • API mutation: Perform a follow-up GET request. Independence matters more than sophistication. If your tool reports success based on its own internal flags, your test will repeat the same mistake. For systems that are eventually consistent, use a polling policy. Set your timeouts and intervals based on your product contract. Stop asking only if the agent ran. Ask which independent state proves the work happened. What surface would you trust after your agent says it is done? Source: https://lnkd.in/gsNPKmyC

A system that only works while the right person is around isn’t really a system. I see this more often than people realise. One person knows the login. One developer knows how the site is built. One plugin holds something important together. One form sends to an inbox nobody checks. One automation breaks and suddenly nobody knows what was supposed to happen next. Nothing feels urgent until the person is away, the tool changes, or something stops working. Then the business finds out how dependent it really was. That is the bit I care about. A good digital setup should be understandable, owned properly, easy enough to maintain, and not held together by one person’s memory. Sometimes that means cleaning up the website. Sometimes it means simplifying tools, fixing access, documenting what matters, or removing a dependency nobody questioned because “it’s always worked”. Resilience is not making everything complicated. It is making sure the business can still move when one small thing changes. If your digital setup feels like it depends too heavily on one person, one tool or one fragile workaround, write to me. I’ll take a proper look and tell you where I’d reduce the risk first.

🚨 LATEST UPDATES TO THE SBA SMALL BUSINESS SIZE STANDARDS 🚨 The Wolverine Group created a Size Standard Impact Tool to help analyze and create responses to the SBA, and we have made some updates to make the tool better and easier. The NAICS Size Standard Impact Tool no longer asks you to download anything. When we released it, using the tool meant pulling a custom extract from USAspending — one NAICS code across five years runs to roughly 240,000 rows — and dropping the ZIP into your browser. It worked, but it put twenty minutes of setup in front of a five-minute question. That step is gone. Now, type the codes you compete in, and the analysis runs. The reason is a collaboration with MakeGov, whose Tango API now supplies the federal award data behind the tool, already resolved from contracting offices to parent companies. Entity resolution is the genuinely hard part of this analysis: a large firm operating under a dozen identifiers looks mid-sized under any one of them, and a tool that misses the connection returns a plausible list of the wrong companies. Tango handles it upstream. My thanks to Dave Zvenyach and the MakeGov team — this is a better tool than the one I could have built alone. It is also more accurate. The prepared data knows each company's whole federal book, so it applies the size test the way 13 CFR 121.104 does — total receipts from all sources, not just the codes you happened to pull. If you still need a scope, a date range, or a reporting-delay cutoff the prepared data does not cover, the USAspending upload route is still there, still reads the file entirely in your browser, and still uploads nothing. Why this matters right now: SBA estimates its proposed rule would newly qualify 37,002 firms that already hold federal contracts — 105,655 contracts worth more than $71 billion in FY 2025. Those firms arrive in the set-aside pool with the capture infrastructure they built competing in the open market. The national figure tells you nothing about your codes. This tells you which specific firms, how much work they hold, and whether they have ever had to win a set-aside. The tool then drafts a comment to the SBA built on your own figures. It takes no position on the rule — you choose oppose, support, support with modifications, or submit the data alone, and the measurement section reads the same either way. Free, no account, no sign-in. The user guide has been rewritten to match. Comments on Docket SBA-2026-0199 close on 21 September 2026. https://lnkd.in/eCeYng4g #FederalContracting #SmallBusiness #SBA #GovCon #Acquisition

If you've ever asked IT to approve a browser extension, you know how that conversation goes. Assessors keep telling us the same thing. The application sits in Optible, their pipeline sits in the CRM or grants system they already use, and the day goes on copying between two tabs. So we built an extension that puts the assessment panel right beside the record in their own system. Open the record, click the icon, score it there, and it's in Optible straight away. Most of the thinking went into making it something a security team can approve without a long meeting. The easy way to build one of these is to let the extension read the page it's sitting on. That hands it access to everything on every page of your system, for everyone who installs it. If you're moving public money, that's a hard thing to sign off, and fair enough. Ours reads the address bar and nothing else. The web address already identifies the record, so there was nothing on the page worth asking for. It's built to work with any CRM or grants management system.