Dazza Greenwood works on legal technology and computational law, including how organizations evaluate AI systems and govern their use in professional work. Their work includes consulting, research, teaching, writing, and community events.
Most useful for advising legal and professional services firms on internal AI evaluation and governance programs.
Best for
Established legal software or professional services firms needing evaluation, governance, and practitioner sessions.
Not for
Paid social endorsement campaigns promoting AI products with benchmark superiority claims or reach targets.
Key takeaways
Generic benchmarks miss an organization's edge cases, terminology, and own standards.
An audit of 88 commercial AI products found roughly 40 percent with instructions against users.
Professionals stay accountable when vendor system prompts are hidden and control instructions.
Check first: Confirm audience includes employer type, job function, and seniority data, not only follower count.
Main ideas
Generic AI benchmarks can't tell you whether a tool works in your organization · Professionals remain accountable when vendor instructions are hidden · Learn to use legal AI agents within explicit authority boundaries
Dazza Greenwood works at the intersection of legal technology, computational law, and AI in professional work.
Their focus includes how organizations can test whether AI meets their needs and how professionals can stay accountable when vendors control instructions users cannot see.
They also teach and convene people around legal innovation, including workshops on AI agents and law and hands-on community events.
Greenwood connects abstract questions about AI governance to concrete practices.
In one post, they argue that general benchmarks miss an organization’s own standards, and point to custom evaluations that measure quality, cost, and risk.
They cite Lake Merritt and shareable, versioned “eval packs” as ways to put that approach into practice.
In another, an audit of 88 commercial AI products—reported as finding at least one instruction working against user interests in roughly 40 percent—grounds a question about professional responsibility and hidden vendor instructions.
The combination of evaluation, accountability, and practical teaching helps readers move from broad AI claims to decisions about specific tools and workflows.
Dazza Greenwood argues that AI in legal and professional work must be measured, governed, and used deliberately—not trusted by default.
Read more
Their posts return to three questions: Does an AI system meet your standards?
Who is accountable when a vendor's hidden instructions control the answer?
How should professionals learn to work with agents that have real authority boundaries?
Greenwood explores these questions through tools, published audits, workshops, and an informal practitioner community.
Generic AI benchmarks can't tell you whether a tool works in your organization
Greenwood argues that benchmark scores describe performance in abstract scenarios but miss an organization's edge cases, terminology, and quality standards.
Read more
The gap between a benchmark score and real-world performance can damage trust.
They propose systematically evaluating specific products and workflows, and treat the ability to define good output in objectively checkable terms as a core executive and governance function—not a technical specialty.
In their view, anyone who can explain good work to a human employee can write a meaningful AI evaluation.
Greenwood released Lake Merritt, an open-source platform for building custom evaluations, in an early public beta aimed at business, legal, and other non-technical leaders.
Quick-start exercises can be tried in minutes, while “eval packs” let teams version and share their evaluation approaches internally or with a wider community.
The platform also supports advanced workflows such as OpenTelemetry and multi-agent system evaluations.
Greenwood says their consulting work has shifted from building and improving AI applications to evaluating whether they meet quality thresholds and stay under cost and risk ceilings.
Artificial Lawyer covered the launch.
This post shifts the commercial emphasis from building AI applications to auditing them, sharpening the governance argument: the value is measurement and oversight, not development.
Professionals remain accountable when vendor instructions are hidden
Greenwood argues that lawyers and other professionals remain responsible for their work, even when an AI vendor can insert higher-priority instructions—hidden system prompts that professionals cannot see, change, or know have changed.
Read more
A carefully written user prompt may not control the answer, so customization alone offers no real assurance. They frame system-prompt assurance as a missing link between professional responsibility and vendor accountability: customization is not governance.
As a co-author of a new AISPA paper, Greenwood reports an audit of 88 commercial AI products that found roughly 40 percent contained at least one instruction working against user interests. A companion blog post applies the finding to professional practice, arguing that meaningful professional responsibility must account for a methodology partly hidden inside the vendor's product.
Learn to use legal AI agents within explicit authority boundaries
Greenwood argues that practitioners should work through the transformation in legal work hands-on, not just hear about it.
Read more
The teaching program they convene covers authority boundaries, legal research, agent skills, HTMLaw, AI-native professional services, and new forms of legal practice.
It moves from talks and demonstrations to team sessions, where participants bring their own agents, form teams, and produce shared outputs in live human/agent collaboration rooms.
Limits on agent authority lead the agenda.
The law.MIT.edu 2026 Summer Intensive takes place online on August 20 and is free to attend.
Speakers include Alex 'Sandy' Pentland and Roland Vogl; the main program features talks, demonstrations, and hands-on exercises.
Registered participants who opt into the exercises or team projects can use Descrybe, a legal research and intelligence platform for U.S. primary law, free for a limited time, grounding the agent work in source-based research.
Greenwood says they are wrapping up active research and teaching at MIT and moving the work through ComputationalLaw.org, Stanford CodeX, and Stanford HAI's Digital Economy Lab. The MIT Computational Law Report will continue under Stanford, with Bryan Wilson as editor in chief, while the CIVICS.Com consulting service stays unchanged.
Practitioner community and informal prompt-sharing are part of the method
Greenwood's community work makes the case that legal professionals learn generative AI through recurring, agenda-free gatherings where people share prompts and hands-on applications over drinks.
Read more
The low-stakes format is open to anyone, including the AI-curious, making it easier to compare real-world practice than to listen to a talk.
The Bay Area Legal Hackers Happy Hour, co-organized by Greenwood and Diana J.
Stern, held its last 2025 event on December 4 at Savoy Tivoli in San Francisco's North Beach.
The series passes to new co-organizers Joachim and Anna for 2026.
Attendance is free and open to all; the invitation describes an evening of mingling, prompt-sharing, and practical generative AI use.
Greenwood is stepping back from running the series, handing organizing duties to new co-organizers even as their other convening work moves to Stanford-affiliated venues.
Answers "why don't benchmark scores tell me whether an AI tool works for my organization?" and names a starting method and tool.
Read more
In a weekend long-read post, Dazza Greenwood argues that generic AI benchmarks miss an organization’s edge cases, terminology, and standards.
Writing objectively evaluable quality criteria, they say, is now an executive and governance function.
The post introduces Lake Merritt, an open-source platform for building custom evaluations, including “eval packs” that can be versioned and shared across an organization or with a wider community.
They also write that consulting work has shifted from building or improving applications to evaluating them against quality thresholds, cost, and risk ceilings, and point readers to further context from Artificial Lawyer.
For brands planning a campaign
A practical anchor for a legal-tech or enterprise AI audience, supporting a written or speaking collaboration on evaluation methods and AI governance rather than generic benchmark claims.
Answers "what does this person teach and convene, and where is the work based now?" in one post.
Read more
Greenwood announces the law.MIT.edu 2026 Summer Intensive, a free online workshop on AI agents and law held August 20, with speakers including Alex 'Sandy' Pentland and Roland Vogl.
The program covers authority boundaries, legal research, agent skills, HTMLaw, AI-native professional services, and new forms of legal practice.
The afternoon adds team-based “Interlateral” sessions where participants bring their own agents and practice live human/agent collaboration.
Greenwood also says they are wrapping up active research and teaching at MIT and moving that work forward through ComputationalLaw.org, Stanford CodeX, and Stanford HAI's Digital Economy Lab, while the CIVICS.Com consulting service stays unchanged.
The MIT Computational Law Report also continues under Stanford, with Bryan Wilson as editor in chief.
For brands planning a campaign
A natural anchor for a workshop, webinar, or conference collaboration on AI agents in legal and professional work, with MIT and Stanford networks as the relevant audience.
Answers "what is this person's specific argument about AI vendor accountability and the instructions users cannot see?" with a concrete audit result.
Read more
Greenwood describes co-authoring the new AISPA paper and writing a short blog post about hidden vendor system prompts.
Their argument: lawyers and other professionals remain accountable for the work they deliver, yet an AI vendor may insert higher-priority instructions that the professional cannot see, change, or even know have changed.
As a result, a carefully written prompt may not be the controlling instruction.
Greenwood reports an audit of 88 commercial AI products that found roughly 40 percent contained at least one instruction working against user interests.
They conclude that customization is not governance and that system-prompt assurance may be a missing bridge between professional responsibility and vendor accountability.
For brands planning a campaign
Strong source for an AI governance or professional-responsibility collaboration aimed at lawyers and professional-services firms, particularly on vendor oversight and audit.
Answers "what community does this person build, and what does it look like in person?" with a clear event description.
Read more
Greenwood co-organizes the Bay Area Legal Hackers Happy Hour with Diana J.
Stern.
The December 4, 2025 event at Savoy Tivoli in San Francisco's North Beach is described as the last of 2025 and a handoff to new co-organizers Joachim and Anna for 2026.
The monthly series is free and open to all, with no formal presentations.
Legal professionals and technologists mingle, share AI prompts, and work through hands-on generative AI uses over drinks.
For brands planning a campaign
Relevant for event sponsorship or community-facing activity in the Bay Area legal tech scene; the series is free, informal, and recurring.
Evaluating AI systems against an organization's own standards
In a long-form post, Dazza Greenwood argued that generic AI benchmarks miss an organization's edge cases, terminology, and quality standards.
Read more
Defining what good AI output looks like in a way that can be objectively evaluated, they said, is now a function of executive leadership and governance.
The post points to Lake Merritt, an open-source evaluation platform in early public beta, including "eval packs" that can be versioned and shared across a team.
It also notes that their consulting work has shifted from building and improving AI applications toward evaluating whether they meet quality, cost, and risk thresholds.
Long-form written explainer paired with a released tool and a walkthrough for non-technical leaders.
For brands planning a campaign
Useful for explaining AI measurement to business and legal audiences without a technical deep dive; the post offers a ready hook for evaluation, quality-assurance, or model-risk messaging.
Greenwood is a co-author of an AISPA paper that audited 88 commercial AI products and reported that roughly 40% contained at least one instruction working against user interests.
Read more
In a companion blog post, they argue that professionals remain accountable for AI-assisted work while vendors can change hidden system prompts professionals cannot see. They frame system-prompt assurance as a missing bridge between professional responsibility and vendor accountability, and say customization is not governance.
Research paper plus a short companion blog post translating the findings for professional practice.
The post reports the study's headline figure and framing rather than the full audit methodology, so the paper itself would be the source for detailed claims.
For brands planning a campaign
Strong fit for governance, compliance, and professional-responsibility angles; the audit statistic is a concrete, quotable finding for legal tech and enterprise buyers.
Greenwood is running the law.MIT.edu 2026 Summer Intensive, a free hands-on workshop on AI agents and law, with sessions on authority boundaries, legal research, agent skills, HTMLaw, AI-native professional services, and new forms of legal practice, plus team-based human/agent collaboration sessions where participants bring their own agents.
Read more
They describe the workshop as a milestone while wrapping up active research and teaching at MIT and moving the work to ComputationalLaw.org, Stanford CodeX, and Stanford's Digital Economy Lab.
Free applied workshop combining talks, demonstrations, hands-on exercises, and team projects.
For brands planning a campaign
A venue and topic fit for legal AI vendors looking to reach practitioners through hands-on education rather than a keynote; the agenda names concrete areas a partner could align with.
Greenwood co-organizes the Bay Area Legal Hackers Happy Hour with Diana J.
Read more
Stern, a free monthly gathering where legal professionals and technologists share AI prompts and practical uses of generative AI in an agenda-free social format. The December 2025 post announces the last event of the year and the handoff to new co-organizers Joachim and Anna for 2026.
Recurring informal meetup rather than a presentation format.
For brands planning a campaign
Offers a recurring, low-cost touchpoint for reaching a legal-tech practitioner community in the Bay Area; the agenda-free format suits informal product demos or prompt-sharing sponsorships.
Greenwood posts live from industry gatherings, including a Harvey session with president Gabriel Perez in conversation with the Notion CEO, and an invitation to connect at Relativity Fest 2025 in Chicago.
Read more
The posts mix brief event observations with open calls for questions and meetups.
Short, in-the-moment social posts rather than analysis.
These are brief event notes; they show presence and access, not sustained analysis of the sessions.
For brands planning a campaign
Suits event-based collaborations and conference visibility; the call-and-response format can carry a brand question to a named speaker or drive meetup attendance.
Dazza Greenwood's posts describe research, teaching, and community affiliations more than commercial sponsorships.
Read more
They state their employment at MIT, where they worked on law.MIT.edu and the Computational Law Report, and describe CIVICS.Com as their consulting service, continuing unchanged while other work moves to Stanford. The posts also describe their co-authorship of the AISPA prompt-assurance paper, Descrybe's participation in the law.MIT.edu summer workshop, the open-source evaluation platform they released, and their co-organizer role with Bay Area Legal Hackers.
MIT
Employment
Greenwood says they are wrapping up active research and teaching work at MIT, where they started in 1997, and cites the law.MIT.edu program and the Computational Law Report as work they have carried on since 2019.
Read more
The post adds that the Computational Law Report will continue under Stanford, with Bryan Wilson as editor in chief.
Greenwood describes the CIVICS.Com consulting service as staying constant while other work moves to Stanford, and addresses existing clients directly about the change.
The post states that after MIT, related research, writing, and convening will continue through Stanford Law's CodeX, the Stanford Center for Legal Informatics.
Descrybe is featured as a legal research resource for the law.MIT.edu 2026 Summer Intensive, making its tools free for a limited time to registered participants who opt into hands-on exercises or team projects.
Read more
The same post announces the workshop and the offer.
Greenwood says they released Lake Merritt, an open-source platform for building and running custom AI evaluations, with eval packs that can be versioned and shared.
Read more
It is described as an early public beta aimed at non-technical business and legal leaders.
Greenwood is a co-author of the new AISPA paper, an audit of 88 commercial AI products that found roughly 40% contained at least one instruction working against user interests.
Read more
They also wrote a companion blog post about its implications for professional responsibility.
Greenwood posted from a Harvey session where its president, Gabriel Perez, walked through the company's history and basic approaches in conversation with the CEO of Notion.
Start with the long-form post on why generic AI benchmarks miss an organization's own edge cases and standards.
Read more
It introduces Lake Merritt, an open-source evaluation platform, and explains why measuring AI quality has become a governance job.
The AISPA post connects that work to accountability, arguing that professionals remain responsible for AI-assisted work even when a vendor's hidden system prompt controls the answer.
For a teaching-focused view, the workshop announcement outlines a free hands-on law.MIT.edu summer intensive on AI agents, authority boundaries, and legal research.
The Bay Area Legal Hackers post shows the recurring informal meetup where legal professionals and technologists share prompts and practical generative AI use.
Two shorter posts from Relativity Fest 2025 and a Harvey session show where Greenwood appears in the legal tech world.
For a different angle, follow Misty Murray for hands-on legal software adoption, Tom McMakin for consulting and business-development practice, or Shakiva Pierre for a real estate law practice built in public.
Both advise legal professionals on adopting technology, publish practical explainers about AI and software in law practice, and run consulting practices serving law firms.
Read more
Each teaches through a concrete problem: Murray uses post-case evaluation with Microsoft Planner, Excel, and Copilot, while Greenwood covers custom AI evaluations and hidden system prompts.
Murray focuses on hands-on Microsoft 365 and Copilot implementation for law firm operations and profitability; Greenwood works on AI evaluation, governance, vendor accountability, and computational law.
Both are independent consultants who publish short, firsthand lessons from client work and run practical workshops.
Read more
Each narrates a real client situation and draws a transferable lesson: McMakin from a business-development call, Greenwood from an evaluation or audit finding.
McMakin writes about business development, account growth, and the consulting business itself; Greenwood focuses on AI evaluation, system-prompt auditing, and changes to legal and professional work.
Both are legal professionals who use their posts to explain their work, document professional development, and share community work in plain language rather than formal marketing.
Read more
They write for fellow practitioners and prospective clients alike.
Pierre's posts describe building a New York real estate law practice across landlord-tenant, transactions, and litigation; Greenwood focuses on AI evaluation, AI governance, and computational law. They address different subjects and audience decisions.
The only measured audience figure mentioned here is a single-channel reach number: Dazza Greenwood's LinkedIn account showed 15,491 followers on 11 October 2026.
Read more
That tells a buyer how many people are connected to one account; it does not say who they are.
For an AI governance and evaluation advisory role with a possible content or speaking collaboration, the practical question is whether those followers include the compliance, legal-ops and product leaders who would approve that kind of engagement, and the posts do not say whether they do.
The posts do show what the audience is being drawn toward.
They cover an applied AI-agents-and-law workshop at law.MIT.edu, the Bay Area Legal Hackers happy hour Dazza Greenwood co-organises, Relativity Fest, and a Harvey fireside with its president.
That pattern points to a legal technology, eDiscovery and computational-law readership.
It is an inference from the subjects that draw engagement, not a measured composition — none of the posts report who registered, attended or follows the account, and the posts do not report engagement counts (likes, comments, reposts) either.
Concrete next step: ask for either a recent audience survey or a platform analytics report covering employer type, job function and seniority.
A follower total on its own cannot support a claim that this audience contains the governance and evaluation decision-makers needed for the advisory role.
If no survey exists, the honest placement is reach into a legal-tech-aware audience, with audience fit for the advisory role unverified.
No sponsorship packages, media kits, rate cards, or disclosed brand-partner placements appear in the material reviewed.
Read more
Dazza Greenwood does state that their consulting service, CIVICS.Com, continues unchanged through the transition from MIT to Stanford, with work shifted toward evaluating whether AI applications meet quality, cost, and risk thresholds.
They also offer free public teaching and community formats: the August 2026 law.MIT.edu Summer Intensive workshop and the Bay Area Legal Hackers happy hour were both open, free to attend, and framed as community and hands-on practice rather than sponsored placements.
For a legal tech company, the practical starting point is an advisory, evaluation, or speaking engagement built around their current research agenda rather than a pre-published advertising slot.
Their posts establish that the consulting service exists and has moved into evaluation work; ask directly about pricing, exclusivity, lead times, and a partnership contact.
Applied AI evaluation and advisory consulting
Buyer requirement
A buyer should bring specific workflows, products, or AI applications to assess, plus a plain-language statement of what acceptable output looks like.
Dazza Greenwood describes the work as custom evaluations that test a client's own edge cases, terminology, and standards rather than generic benchmarks.
They say articulating evaluable quality is now an executive and governance responsibility.
Their posts do not specify rates, scope limits, contracting terms, or turnaround; confirm those directly before planning.
Workshops, talks, and community sessions
Buyer requirement
Dazza Greenwood's published teaching events were free and open to all, with registration handled through event pages.
The August 2026 workshop included optional after-session team exercises where participants brought their own AI agents.
Confirm directly whether a sponsored, private, or client-specific version is available; the posts show only open public programming.
Brief: an established legal-software or professional-services company wants an outside adviser to stand up an internal AI evaluation and governance program for legal workflows and co-host a practitioner session.
Read more
Dazza Greenwood is a strong fit.
Their published material says the service has moved from building and improving AI applications to evaluating whether they meet quality, cost and risk thresholds.
It names CIVICS.Com as a consulting service that continues through their move from MIT to ComputationalLaw.org, Stanford CodeX and Stanford HAI's Digital Economy Lab.
They have a concrete position to build a program around: generic benchmarks miss an organization's edge cases and terminology; defining evaluable quality is an executive and governance job; and hidden vendor system prompts mean a professional's own instructions may not be the controlling ones.
A paper they co-authored audited 88 commercial AI products and reported that roughly 40% contained at least one instruction working against user interests.
They also convene and teach this material.
The law.MIT.edu 2026 Summer Intensive was a free workshop on AI agents and law, with hands-on exercises and afternoon team sessions where participants brought their own agents.
It covered authority boundaries, legal research, agent skills and AI-native professional services.
A co-hosted session in that format reaches a legal-tech and computational-law readership and produces practitioner-level output rather than a keynote.
Fit is strongest when the buyer brings real workflows and edge cases; it is weak if the buyer wants a generic benchmark endorsement.
When to choose a specialist
An AI vendor wants a paid social endorsement campaign: a scripted post or short video promoting the product, affiliate or tracked links, a follower-reach target and a messaging angle built on claimed benchmark superiority.
Read more
Dazza Greenwood's posts do not support that shape; some argue against it.
The posts disclose no sponsorship packages, media kits, rate cards or brand placements.
The public teaching and community formats described—the law.MIT.edu workshop and the Bay Area Legal Hackers Happy Hour—are free and open, not sponsored slots.
The audience case is thin for a reach-based buy: the only measured figure is a single LinkedIn follower count, with no data on employer, function or seniority, and no engagement counts.
More importantly, the published argument cuts the other way.
Their writing treats generic benchmark scores as a misleading signal and vendor-controlled hidden instructions as a governance problem, so a scripted claim that a product wins on benchmarks would contradict the position the buyer is paying to borrow.
A vendor with a paid-placement budget and a reach KPI should look at creators who publish sponsorship terms rather than try to convert a consulting and community practice into one.
A worked campaign example
Angle: "Your benchmark score is not proof of fit — build three evals for your own processes."
The vendor supplies one real product or process, three genuine edge cases where it currently underperforms or is unproven, and its own internal definition of acceptable output.
Greenwood delivers a private 60-minute session for the legal, compliance and product leads, modeled on the free workshop format: a short framing talk on why benchmark scores travel badly into a specific organization, a demonstration of building custom evaluations with versioned eval packs, and a hands-on block where the client's team writes and runs its first eval against its own edge cases.
A follow-on written piece turns the session into a shareable internal case study; a public version can be pitched to a legal-tech community audience where Greenwood already has standing.
This angle draws on an argument already published and aimed explicitly at non-technical business and legal leaders, and on a workshop format Greenwood actually runs.
The client keeps a working evaluation, giving the collaboration a measurable result instead of an impression count.
Measure whether the team produced evals that pass its own thresholds, whether a named owner carries the eval practice forward, session completion and qualified follow-up conversations.
Before planning, confirm fee and scope, whether any private or paid version of the public workshop format exists, turnaround, exclusivity and the exact affiliation wording permitted for ComputationalLaw.org, Stanford CodeX and Stanford HAI's Digital Economy Lab.
If audience reach is part of the value, ask for platform analytics or an audience survey covering employer type, job function and seniority; the single follower figure cannot show whether the account reaches the governance and evaluation decision-makers this campaign targets.
Worth switching to if the campaign shifts from AI governance to New York real estate law—residential and commercial landlord-tenant matters, transactions, and litigation—or to a session for real estate attorneys on building and running a practice.
Read more
Pierre’s published work is squarely in that area: she describes launching the Law Office of Shakiva S.
Pierre, Esq.
PLLC after nearly a decade in New York real estate law and time at two New York City real estate firms, serving as a chair of the Real Estate Law Section of the Metropolitan Black Bar Association, and writing about development, networking, and community work as part of growing a firm.
For buyers targeting small-firm and real estate practitioners with a session on practice-building rather than AI, she offers subject matter Greenwood does not cover.
Pierre’s two posts cover career milestones, practice areas, and event attendance, not AI, evaluation, or governance.
Choosing her for an AI evaluation and governance campaign means paying for a subject she has not published on; you’ll need a separate source for the evaluation and governance substance.
She publishes no sponsorship terms, rate card, or speaking formats, so ask her for those before discussing a collaboration.
A better choice when the need is go-to-market and account-expansion training for the professionals who would carry an AI program forward, rather than the governance framework itself.
Read more
McMakin’s writing covers exactly that: a business-development training proposal to a healthcare consulting firm focused on building key accounts and expanding work inside existing clients, and a candid account of a call where asking what a client was trying to get done—rather than pitching—surfaced a thought-leadership need. If the legal tech company’s problem is that its law-firm clients buy one product and never expand, this is the closest published match among the three alternatives.
The posts cover business-development coaching for consultancies, not legal AI.
They do not address evaluation, hidden vendor instructions, or legal processes, so a campaign about Greenwood would trade governance and evaluation expertise for sales coaching; you’d need a separate subject-matter expert for the evaluations.
One of the two posts is McMakin passing along another consultant’s workshop promotion, not their own analysis.
Ask for original analysis if that is important to the campaign.
Consider Murray when the buyer wants implementation inside law firms rather than an evaluation methodology—teaching legal teams to get more out of tools they already pay for.
Read more
The posts make that concrete: a Copilot blog series arguing that Microsoft 365 (Planner, SharePoint, OneNote, Lists, Excel) should be used with intention to see how a team really works and where processes stall, and a case-autopsy tip using Planner, Excel, and Copilot to judge whether a contingency case was worth it and refine intake criteria. That covers the operational side of a campaign seeking measurable processes, adoption inside a firm, and process sustainability, at a more hands-on, tool-specific level than Greenwood’s published material.
The scope is narrower.
Everything reviewed sits inside the Microsoft 365 and Copilot ecosystem and is framed as tips and a blog series for Arrow Consultants, so the buyer gets platform-level operating advice rather than a method for judging whether an AI system clears a firm’s own quality, cost, and risk thresholds.
Ask for examples of independent evaluation work or convening and teaching if those are campaign requirements.
Dazza Greenwood published six LinkedIn posts between late 2025 and 2026: a long-form argument for custom AI evaluation and the release of the Lake Merritt open-source platform; a companion post on a co-authored AISPA paper auditing hidden system prompts in commercial AI products; the announcement of the free law.MIT.edu 2026 Summer Intensive workshop on AI agents and law; the final 2025 Bay Area Legal Hackers Happy Hour; and two short posts from Relativity Fest 2025 and a Harvey session.
The similar-creator comparisons draw on two posts each from three other legal and consulting professionals — Misty Murray, Tom McMakin and Shakiva Pierre.
A single LinkedIn follower figure is shown for 11 October 2026.
The posts are Greenwood's own summaries and announcements.
The posts here do not provide details about the linked papers, the Lake Merritt software, workshop sessions, recordings, engagement counts, audience composition, sponsorship terms or pricing, and the relationship entries do not give exact post dates.
The posts do not establish a signature book, show or earlier body of work, so this page makes no such claim.
