IIIMPACT Design

06/10/2026

New Anthropic policy: "To ensure we're responsibly deploying Mythos-class models, we are requiring limited data retention and review as part of our safety work. Prompts submitted to, and outputs generated by, Mythos-class models are retained for 30 days for trust and safety purposes, on every platform where these models are offered. "

If you use fable/mythos - they collect your data ...no exceptions even for enterprise partners.

Even if your organization previously negotiated zero-data retention agreements, you are now subject to this new 30-day window.

Advanced power now comes with a new set of data governance trade-offs.

Anthropic is like an abusive parent who buys you the best toys so you just deal with it.

Claude is Anthropic's AI, built for problem solvers. Tackle complex challenges, analyze data, write code, and think through your hardest work.

05/30/2026

Stop looking at cost per API call. Measure the Cost Per Successful Outcome.

Moving faster in the wrong direction, is far more expensive than moving slowing in the right direction.

05/30/2026

When Compute Costs More Than Headcount

We are officially seeing AI compute expenses actively outweigh the personnel costs they were supposed to replace.

The biggest mistake happening right now? Treating AI like a static SaaS license. AI is a variable expense. Poor usage governance recently cost one enterprise $500M in a single month because they didn't cap employee access...but hey you all have to move faster and use "more AI".

The token costs for running AI agents are now exceeding what they were paying the employees they fired.

When the tokens run out, the AI stops. Just stops. No continuity. No workaround. Just a spinning wheel where your workforce used to be.

You fired humans to save money and bought a subscription that bills you into a corner.

The employees you let go knew what to do when things broke.

05/29/2026

Large language models are backward-looking by design. They train entirely on historical data. They do not invent. They synthesize what already exists.

When you rely on an AI model for product strategy or creative direction, you are generating derivative work. You are producing a clone.

Clones do not command a premium.

Every company right now is making a fundamental choice about their market position.

You can use AI to optimize your ability to copy. It is faster and cheaper than ever to replicate what your competitors are doing. If you choose this route, you will permanently play catch-up. You will trade market leadership for operational efficiency.

Or you can prioritize forward-looking human creativity. You can build the concepts that simply do not exist in the training data yet. By the time the rest of the market trains their models on your work, you will already be on the next iteration.

Market Leader > Market copier

05/29/2026

Software engineers are slowly losing the ability to navigate their own codebases.

When developers write code natively, they build a mental model of the architecture. They know where things are, how systems interact, and why certain fragile decisions were made years ago.

When developers transition to simply prompting an AI to generate code, that mental model degrades.

They are outsourcing the implementation, which means they are outsourcing their understanding of the system.The problem surfaces when something breaks. A developer who hasn't written the code no longer knows how to fix it natively.

They are forced to continue prompting their way out of the problem, guessing at solutions because they no longer possess the deep, structural knowledge of how everything is wired together.

We are trading deep expertise for temporary speed.

05/28/2026

The tech industry is walking blindly into a massive pricing vulnerability with AI coding tools.

Developers are rapidly transitioning from writing code to prompting it. As a result, engineering teams are losing their native familiarity with their own software architecture. They are becoming entirely dependent on external AI models to generate, debug, and maintain their codebases.

Uber blew through a 12-month AI budget in exactly four months.

Microsoft watched API costs climb and forced its own engineers to abandon Anthropic's Claude Code, despite those engineers preferring it over Copilot.

If AI providers execute a rug pull and raise token prices, engineering teams will be fuct. They cannot simply revert to writing and maintaining the code manually.

The developers will have already lost the deep institutional knowledge required to navigate their own systems natively.

When your team relies entirely on an external AI model to maintain your product, you are no longer paying for a software tool. You are paying rent on your own codebase.

The AI landlords WILL raise token prices because they need to appease shareholders when they go public.

When this happens, and you've let go a significant portion of your team that had institutional knowledge, you competitors that retained their experienced teams with proper AI governance... they will be able to surpass your overworked, short-staffed design/dev teams with ADHD trying to manage 5 agents to just keep up.

05/28/2026

AI coding tools are creating a massive bottleneck in software development.

Development teams are outputting code significantly faster right now. But faster output does not mean better output. All that AI-generated code is simply moving the bottleneck downstream and completely overloading QA teams.

Coding is only a fraction of the work required to ship software. Code review, testing, security scanning, integration is still highly manual. When engineers use AI to generate massive code diffs at an unprecedented pace, QA teams are buried under mountains of code they now have to verify line by line.

We are treating software development like a factory where the assembly line is running at 10x speed, but there's still only one inspector at the end of the belt.

The result isn't faster software delivery. It is a massive backlog and a high risk of broken code making it to production.

I've mentioned before, we were driving a bus and now have given keys (AI) to an F1 race car to a teenager. Sure we can move faster, but we will also be moving in the wrong direction and can crash more often.

05/22/2026

Ever read Through the Looking-Glass? Alice steps into a reversed world, starts running with the Red Queen, and realizes they aren't going anywhere. The Queen tells her: "It takes all the running you can do, to keep in the same place."

Fast forward to today’s tech landscape. Does that sound familiar?

In our newest episode of the Make an Impact Podcast, Makoto and Brinley tackle the Red Queen Problem. We discuss the unspoken reality of AI burnout, the rapid decay of skill relevance, and why just "working hard" doesn't translate to career advancement anymore.

If you feel like you are constantly sprinting just to avoid falling behind, this episode is for you. We explore the choice between augmentation and abdication, and share practical strategies to regulate your energy and focus on the skills that actually matter.

Catch the full trailer and episode here: https://youtu.be/uqN-0qFy0sY

05/20/2026

If you have any private repos with plan text secrets or sensitive documents/arch, immediately rotate your secrets.


We are investigating unauthorized access to GitHub’s internal repositories. While we currently have no evidence of impact to customer information stored outside of GitHub’s internal repositories (such as our customers’ enterprises, organizations, and repositories), we are closely monitoring our infrastructure for follow-on activity.

05/20/2026

AI is really big data sets, lots of compute, LLMs mushed together, datasets by their very nature is backward looking, creativity is forward looking.

Because AI is backward-looking (trained on what already exists) and creativity is forward-looking (inventing what doesn't exist yet), it can only produce derivative work.

Clones don't sell.

- Strauss Zelnick, CEO of Take-Two