Marcus

@utf9k.net

294 Followers 191 Following

I use computers for work during the day and to write letters to the government at night https://utf9k.net

An email from Iain Lees-Galloway at 6:46 AM It says the following: Kia ora Marcus, You have received this email because you purchased merchandise from the Opportunity Shop earlier this year. This email is only for people who purchased Opportunity Party branded caps. If you did not purchase a cap, please disregard this email. We have received conflicting advice from the Electoral Commission regarding the caps. Previously, they were not considered election advertisements and did not require a "promoter statement" (the text that starts Authorised by..."). However, recent advice has called that into question. Right now, it is unclear to us whether caps require the promotor statement. The cap(s) you purchased do not have the promoter statement. While we clarify this, please keep your cap(s) out of sight. Put them away in your wardrobe or a drawer. Displaying an election advertisement without the promoter statement is a criminal offence and could be referred to the Police.
A photo of a Macbook Pro. An ethernet cable is plugged into it via a USB-C to ethernet adapter. An ethernet cable runs from the laptop behind an LG TV. The TV displays a poorly drawn screen that reads "Marcus's better hotel menu :)" with a single "play content" button below it.
Heading: The Ministry of Justice highlighted differences in how the responsibility relating to obeying the law is worded in citizenship material 31. In the current citizenship by grant application form and in guidance on govt.nz, one of the responsibilities of New Zealand citizenship is listed as 'obey and promote the laws of New Zealand'. The Ministry of Justice has highlighted that there is no legal obligation on citizens to promote New Zealand laws, only to obey them.
A screenshot of a chart titled "Increasing alignment and savings". The time scale shows "FY25/26", "FY26/27", "FY27/28" and "Outyears". There are three arrows (one green and two blue) that say "Alignment and savings". They start faint and get stronger as they flow towards the right.
The box for Patlabor 2 The Movie 4K Remaster Collection sitting on a desk
A screenshot of the Digital Target State diagram. In the centre is a box titled Digital Public Infrastructure which contains generic software concepts such as "Notifications & Messaging System", "Data Dictionary", "Rules Library" and so on.
A photo of Persona 4 Golden for the Playstation Vita. It's signed by David Farrier of Tickled Fame in the bottom right corner.
A still from the TV series Trailer Park Boys where the character Ricky is handing Jim a $100 Canadian bill. The subtitles read "I'm gonna pay you $100 to fuck off"
A photo of Encyclopedia .hack and .hack//Analysis on a white desk
A screenshot of a Written Question to Parliament which says the following: Teanau Tuiono to the Minister for State Owned Enterprises: What advice, reports, briefings, emails, aide memoires, or any other documents has the Minister or any member of their Office seen, if any, about WorkSafe New Zealand safety requirements for fuel storage tanks? Hon Simeon Brown: As I said in Written Parliamentary Questions 10889 (2025), 11147 (2025), 14770 (2025), 18421 (2025), 19559 (2025), 20746 (2025), 22412 (2025), 27507 (2025), 28436 (2025), 30419 (2025), 30991 (2025), 31535 (2025), 32421 (2025), 33193 (2025), 34522 (2025), 35938 (2025), 37794 (2025), 38920 (2025), 39747 (2025), 41640 (2025), 42857 (2025) , 44980 (2025), 4594 (2025), 48408 (2025), 49591 (2025), 50437 (2025), 51143 (2025), 51949 (2025), 52883 (2025), 53446 (2025), 54774 (2025) 55541 (2025), 57127 (2025), 982 (2026), 723 (2026), 4196 (2026), 4299 (2026), 4402 (2026), 4507 (2026), 5769 (2026), 5875 (2026), 6395 (2026), 7771 (2026), 7877 (2026), 8911 (2026), 10970 (2026), 11077 (2026), 9128 (2026), 9247 (2026), 11485 (2026), 11558 (2026), 12455 (2026), 12597 (2026), 13459 (2026), 13609 (2026), 16629 (2026), 17535 (2026), 18356 (2026), 19298 (2026), 24632 (2026), and 27221 (2026) I refer the Member to the answer to Written Parliamentary Question 6125 (2025).
A panel from Chapter 35 of Smoking Behind The Supermarket With You. Sasaki is caught red handed overspending on fish out of jealousy.
A bookshelf with a number of books facing the camera. One is bright red with a smiley face on it.
Phase two of the Inquiry is likely to cover: - Whether patients were properly asked for authorisation before a MMH account was established for them and information was stored in that account. - Whether patients received adequate information about how the portal would be used. - Retention and deletion of information within the portal. - The quality of communications about the breach. - Whether the notifications to OPC and to affected patients met the requirements of the Privacy Act. - Whether the breach caused a disproportionate impact on any group, particularly Northland Māori.
What did John Key say yesterday? Oh, they have issued them with iPads. They have issued them with iPads. That is going to help when you are faced with somebody screaming and ranting, high on P—when you hold up your iPad! I hope that it is on Pokémon GO and you can get the nearest Zubat. Is that what it is going to be? Is that what the answer is to out-of-control crime because of this Government’s failure to resource the police properly? Well, I tell you what: the rest of New Zealand cares. And, of course, the Government has now discovered, because David Farrar has told it, that the rest of New Zealand cares and expects something better.
A table with the title "Principle 8: How will you ensure that the information used is correct?" It says: We rely on the information in the nomination and the letters of support and do not check their accuracy. This is a feature of the current system and we do not propose to change it. As to whether Parata's citation is accurate, an Honours Unit staff member will check the citation for accuracy against the original nomination and letters of support before it is considered by the APH Committee. The same (or similar) prompt will be used in Parata to instruct it to prepare the summary, for consistency and to reduce bias.
A screenshot of a table with the heading "WorkSafe NZ" and the subheading "CoPilot Chat Proof of Concept" The summary reads "Demonstrate the value of introducing CoPilot Chat with two use cases including OIA responses and guidance material" It is in the "Plan & Design" activity stage with Microsoft and other government agencies as their key partners.
A screenshot of a table with the heading "Statistics New Zealand" and the subheading "Price Web Scraping" The summary reads "Web Scraping in our unit is difficult due to websites changing often. In order to fix it currently we go back to our scrapers and manually adjust the HTML selectors and code. The idea is to use a GPT model as a 'processor' to extract price data from HTML body text. This will hypothetically eliminate the need for HTML selectors and specific code" It is in the "Operate & Use" activity stage with OpenAI as their key partner.
A screenshot of a table with the heading "Statistics New Zealand" and the subheading "AI Image Generation for Communications" The summary reads "Need for engaging and illustrative images to improve internal communications and articles. Using AI image generation tools to create visual content for internal communications, articles, and publications" It is in the "Plan & Design" activity stage with Microsoft Copilot as their key partner.
A screenshot of a table with the heading "Reserve Bank of New Zealand" and the subheading "Digital Cash - Public Consulation" The summary reads "The Digital Cash programme has been exploring the high level design options for digital cash, and asking for public feedback on the proposed design features. We received over 18,000 survey responses, and over 500 written responses. This was identified as a good candidate for an AI Proof of Concept (PoC) because of the otherwise heavy manual effort required to analyse the responses and because the nature of the data lent itself well to standard Retrieval-Augmented Generation (RAG) approaches. The AI PoC had the following objectives: Explore AI for analysing Digital Cash Consultation responses and performing sentiment analysis of unstructured datasets. Determine AI's potential to reduce manual effort and uncover new insights. The team approached the PoC by: Using generative AI tools, specifically Microsoft Co-Pilot Studio, to build a chatbot that was deployed as a Microsoft Teams application. Leveraging RAG technique to quickly build an AI application without having to train the AI on submission data." It is in the "Plan & Design" activity stage with Microsoft, Azure and Copilot as their key partners.
A screenshot of a table with the heading "New Zealand Trade and Enterprise" and the subheading "CRM Agent" The summary is redacted under Section 9(2)(ba)(i) of the Official Information Act It is in the "Plan & Design" activity stage with TBC as their key partner. That is not a name, it is to be decided.
A screenshot of a table with the heading "New Zealand Qualifications Authority" and the subheading "Automated text scoring" The summary reads "Use of essay scoring model powered by AI to mark the literacy exam" It is in the "Operate & Use" activity stage with New Data Solutions as their key partner.
A screenshot of a table with the heading "New Zealand Defence Force" and the subheading is redacted under Section 6(b)(i) of the Official Information Act The summary is redacted under Section 6(b)(i) It is in the "Plan & Design" activity stage with redacted key partners.
A screenshot of a table with the heading "New Zealand Defence Force" and the subheading is redacted under Section 6(a) of the Official Information Act The summary is redacted under Section 6(a) It is in the "Plan & Design" activity stage with redacted key partners.
A screenshot of a table with the heading "Nelson City Council" and the subheading "Analysis of submissions" The summary reads "Problem Statement: With limited time to process community feedback on the Long Term Plan, an innovative approach was needed to efficiency review submissions and deliver insights to elected members. AI Solution: Partnering with a local AI specialist, a secure tool was developed to streamline the analysis of public submissions. Using OpenAI technology, the tool processed over 1,500 entries from a submissions database, undertaking sentiment analysis and identifying key themes. While staff reviewed and verified outputs, the AI tool provided structured, auditable data that informed reporting and supported decision-making. The tool also enabled real-time responses to queries during meetings and was later used to draft personalized replies to submitters, reflecting Council decisions. Benefits and Outcomes: Enabled deeper understanding of community feedback to inform decision making; Reduced manual workload, allowing staff to focus on strategic tasks; Designed for reuse across various consultation types; Delivered timely, data-driven insights that improved the quality speed of Concil deliberations" It is in the "Operate & Use" activity stage with AI Factory as a key partner.
A screenshot of a table with the heading "Natural Hazards Commission" and the subheading "Land Assessment Data from Claim File Notes" The summary reads "Analysing file notes for settlement information. The goal is to identify whether claims were settled based on remediation or valuation. This involves using AI to parse through the settlement notes and extract key information, such as the basis of settlement and specific details like square meterage. This use case aims to automate the extraction of structured data from unstructured text in settlement notes." It is in the "Plan & Design" activity stage with no key partners.
A screenshot of a table with the heading "Ministry of Social Development" and the subheading "AI Client case note search" The summary reads "Cases are opened for investigation. Part of the investigation process is reading large volumes of case notes from a client management system. This is a challenge due to the large amount of case notes, as well as the inconsistent structure of case notes. AI can create a summary of a case note that is: Neutral in tone, Consistent in language, highlights important setences, create(s) a narrative sumary. In addition, it can help people find information inside of unstructured case notes. This includes the generation of synthetic case notes as part of the testing process." It is in the "Develop" activity stage with Deloitte as key partners.
A screenshot of a table with the heading "Ministry of Māori Development - Te Puni Kōkiri" and the subheading "Submission review" The summary reads "We are in the early stages of considering how we can use AI to review submissions to various legislation changes, to speed up the processing of public submissions" It is in the "Plan & Design" activity stage with no key partners.
A screenshot of a table with the heading "Ministry of Justice" and the subheading "Copilot trials - OIA agent" The summary reads "Actively exploring and utilising an AI Agent for OIA's to achieve efficiencies and service improvement. This agent speeds up retrieval and collation of responses to OIA and PQ requests to reduce staff effort, improve response speed and accuracy, and lower risk of errors" It is in the "Develop" activity stage with Microsoft and Engage Squared as key partners.
A screenshot of a table with the heading "Ministry of Justice" and the subheading "Copilot for Generating abstracted for publishing judicial decisions" The summary reads "The Ministry publishes approx. 800 Disputes Tribunals decisions annually. Summarised case notes of each decision are manually created which take approx. 20 minutes per document. AI technology is being used to provide a first draft of summarised decisions, before they are fact-checked, edited and published" It is in the "Operate & Use" activity stage with Microsoft and Engage Squared as key partners.
A screenshot of a table with the heading "Ministry of Disabled People" and the subheading "Finding a document on a NZ Government website" The summary reads "I was unable to find the document I needed to access using the search function on the departments own website so asked Copilot to search for it" It is in the "Develop" activity stage with MS Copilot as the key partner.
A screenshot of a table with the heading "Ministry of Disabled People - Whaikaha" and the subtitle "Analysis of NZDS submissions" The summary reads "Using Nvivo and Copilot to theme submissions to the New Zealand Disability Strategy, look at differences across population groups and suggested changes" It is in the "Plan & Design" activity stage with Copilot as their key partner.
A screenshot of a table with the heading "Ministry for the Environment" with the subheading "Waste Applications" The summary reads "Use Copilot to help the waste teams analyse the applications and supporting information under the waste minimisation act. Applications can be lengthy, upwards of 200 pages in addition to 6 or more supporting documents. Our problem is that there is a lot of information to consider, which takes time, which we have a limited supply of. In addition, as these applications are not a priority, we only pick them up when resourcing allows, which means applications can be on our desk for several months. This delay can be frustrating for applicants." It is in the "Plan & Design" activity stage with Deloitte as the key partner.
A screenshot of a table with the heading "Ministry for the Environment" and the subheading "Document redactions" The summary reads "An agent to do redactions (teach it based on every document we redact with meta data on the grounds/reason)" It is in the "Plan & Design" activity stage with Deloitte and Datacom as key partners.
A screenshot of a table with the heading "Ministry for the Environment" and the subtitle "Waste Infrastructure Model" The summary reads "Potential to create a digital twin to help support the investigation of highly varied data (geospatial, images, border data, structured, unstructured). Could be used to road test policy changes, and provide predictions of waste volumes distribution, life of landfills, and candidate sites" It is in the "Plan & Design" activity stage with Deloitte/Datacom as key partners.
A screenshot of a table with the heading "Ministry for Regulation" and the subtitle "All of Government Regulatory system mapping" The summary reads Development of an interactive network diagram depicting the connections between primary legislation, administering agencies, portfolio and secondary legislation (where visible). We used AI to help identify connections between the network entities." It is in the "Develop" activity stage with Microsoft as their key partner.
A screenshot of a table with the heading "Ministry for Ethnic Communities" and the subtitle "Sppech for ethnic communities" The summary reads "Needed to find out information on ethnic communities within NZ - whilst we had data, needed more specific information and this was supplied" It is in the "Operate & Use" activity stage with 'co-pilot' as the key partner.
A screenshot of a table with the heading "Intelligence Community" and the subtitle "Classified" The summary reads "The Intelligence Community identified that they have 10 classified AI use casss" It is in the "(Blank)" activity stage with no listed key partners.
A screenshot of a table with the heading "Institute of Environmental Science and Research" and the subheading "Synthetic Population" The summary reads "We have developed a synthetic population by harvesting data from the IDI and other sources, that is statistically significant at the SA2 level. Using a large language model (LLM) we developed realistic life diaries for the 5.2 million synthetic kiwis to create dynamic data to support modelling over a range of domains, including the spread of infectious diseases and the impact on health of climate change or any population modelling challenge. This synthetic population is a useful surrogate for real data for research, as it enables us to experiment with a population that resembles reality without risking breaching data privacy" It is in the "Develop" activity stage with the key partners section saying "This syntethic population is developed inhouse with the use of data from numerous sources, including the IDI".
A screenshot of a table with the heading "Inland Revenue Department" and the subtitle "Legal Research Summarisation" The summary reads "Lexis+AI pre-production Beta testing of Generative AI capabilities for Legal research" It is in the "Develop" activity stage with LexisNexis as a key partner.
A screenshot of a table with the heading "Health New Zealand - Te Whatu Ora" with the subheading "Clinical Risk Scores & Clinical Process Support" The summary reads "Machine learning to analyse patient data and predict the likelihood of adverse health outcomes, such as hospital readmission or disease progression. Machine Learning embedded in medical devices to support clinical processes" It is in the "Operate & Use" activity stage with no key partners.
A screenshot of a table with the heading "Financial Markets Authority" and the subheading "Extracting data from PDF files" The summary reads "Utilise AI to extract information from PDF documents into data spreadsheets. We piloted this on financial statements, which are inconsistent in nature so a good test of AI capability" It is in the "Develop" activity stage with Microsoft Power Automate as their key partner.
A screenshot of a table with the heading "Department of the Prime Minister and Cabinet" and the subheading "'Paerata' - a locally hosted generative AI tool using Chat GPT" The summary reads "At DPMC we currently use 'Paerata' - a locally hosted garden-walled generative AI tool build for use in-house using Chat GPT, operating within the Azure environment. All staff at DPMC have access to this too, which is the only approved tool for staff use. Paerata does not learn from the prompts and queries inputted, and each query can be deleted by users and is not cached. Personal information should not be entered into Paerata. A separate application can be made by way of a Privacy Impact Assessment for that type of use case. Staff are able to input DPMC information to to and including RESTRICTED. Staff from all functions are currently using Paerata for a range of tasks, including preparing first drafst of briefings, strategy and corporate business documents, and communications. They are also finding it useful for queries on terms, some research and common questions that would usually be entered into search engines. Staff have been given good guidance about the limits and risks of 'hallucinations' and advice on the safe and responsible use of AI. This is underpinned by a clear governance structure on AI, a staff-led AI working group and an organisational GenAI policy approved by our Executive Leadership Team." It is in the "Operate & Use" activity stage with Treasury/CASS as the key partner.
A screenshot of a table with the heading "Department of Internal Affairs" and the subheading "Identification of problematic language in catalogue records" The summary reads "Use of LLMs to identify potentially harmful language in the catalogue description of collections / holdings of National Library. This has been used to reduce the number of items staff need to look at and verify. There would be too many descriptions for a staff to look at without having AI reduce the number to verify." It is in the "Operate & Use" activity stage with no key partners.
A screenshot of a table with the heading "Department of Internal Affairs" and the subheading "AI redaction tool" The summary reads "To experiment with the use of AI for redacting content from large volumes of information that would be publicly released" It was in the "Plan & Design" activity stage with Microsoft and Internal Business Partners. Below the table is red text that reads "The Department of Internal Affairs notes this use case did not progress."
A screenshot of a table with the heading "Department of Corrections" and the subheading "Productivity Improvements" The summary reads "Use if AI to reduce repetitive work allowing more time for meaningful work" It is in the "Develop" activity stage with Datacom and Microsoft as key partners.
A screenshot of a table with the heading "Bay of Plenty Regional Council - Toi Moana" and the subheading "AI-Assisted Public Submission Analysis" The summary reads "This use case proposes leveraging artificial intelligence to streamline the analysis of public submissions received in response to consents, policies, or planning documents. Whether feedback is legally mandated or voluntarily solicited, the AI system would apply a consistent workflow regardless of submission volume, scaling from 10 to 1,000+ responses. Key AI functions include (1) automatically classifying submissions by stance (support, oppose, neutral) and extracting rationale; (2) identifying and grouping feedback into thematic categories aligned with the source document; and (3) generating a structured recommendation report to support decision making" It is in the "Plan & Design" activity stage with Microsoft and Spark as key partners.
A screenshot of a table with the heading "Auckland Transport" and the subheading "PDP Agent" The summary reads "The PDP Agent is an AI-powered assistant designed to help employees set and manage their performance development plans (PDP). It uses job descriptions and team objectives to suggest relevant goals and deliverables and represents these in a structured format. The agent also incorporates the SMART framework to ensure goals are specific, measurable, achievable, relevant, and time-bound" It is in the "Operate & Use" activity stage with Microsoft as a key partner.
A screenshot of a table with the heading "Accident Compensation Corporation" and the subheading "AI supported privacy checking" The summary reads "Understanding how AI can be used to identify various information types in documents that requires redaction from a staff member." It is in the "Plan & Design" activity stage with Datacom as a key partner.
A photo of a smiling turtle upside down. Its underside is a compact disc. Text below the turtle reads "Program material is recorded on the other side of this disc"
A screenshot of a text file containing many links. A "Disallow: /" clause is added for a number of "Speciality bots". Anthropic (Claude) and OpenAI (ChatGPT)'s various user agents are listed
4. Privacy Risk Assessment Automation AI can support agencies in conducting Privacy Impact Assessments (PIAs) by automating parts of the risk analysis process, especially for standardised or repeatable scenarios. This could be offered as a tool or service by OPC to regulated entities.
A heading reads "OPC has a no AI policy at present but is open to ways in which AI could be deployed" Paragraph text reads "OPC currently has a 'No AI' policy in place at present. This has been a cautious approach for good reason, given the need to uphold the highest standards of privacy and public trust. However, the public is already using AI to converse with OPC, and it is enabling them to do so fast. AI is already-". The rest of cut off.
A screenshot of a section of the report. The heading "Psychosocial risk" is visible with the rest of the it blacked out under S9(2)(g)(ii). This is followed by a large black box and then a bit of text is visible which reads "Nonetheless, the psychosocial risk inherent in the role as currently structured underscores the importance of our previous recommendation". This is then followed by more redactions.
A screenshot of a Google Search with the query "homies don't share utm links". The Google AI overview for the search results says "You are spot on."
A screenshot of an hCapture featuring 9 images with the question "Tap objects that can reflect light"
A screenshot of an NZ Police PDF file open in Safari. Two slides are visible titled "Cost Pressures & Fiscal Risks". Both slides are covered entirely in black boxes.
A screenshot of a table containing various metadata such as the folder that an issued OIA lived in, the template name, PDF creation tool and so on
A photo of a TV saying "Agentic AI for Government Grants"
A photo of a dashboard showing a question about a pharmeceutical company using AI to identify potential drug candidates. There are four options on how to address the research team wanting to fast track clinical trials.