The Huddle: Issue 09

A biweekly look at what's changing in AI, ways to apply it today, resources you didn't know you had, and sharing wins.

How to keep up without drowning

Sometimes the people who feel most behind are usually the ones trying hardest to keep up with all of it. That’s true beyond the world of AI, but especially in the whirlwind of change… Trying to keep up with every model launch and feature drop can be exhausting. It’s like a movie buff feeling like they have to consume every film that comes out every week - it’s constant! We’d rather help you find the award winning films and the off the beaten path classics so you can restore some sanity.

The goal of this issue of The Huddle: get more out of the tools you already pay for, cover the handful of things from the last two weeks that earn your attention, and point you toward a way to build the habit instead of refreshing a feed.

WHY YOUR AI GETS DUMBER (AND MORE EXPENSIVE)

Any user of AI beyond basic chatbot has felt this at one point. You start a chat and the AI is sharp, giving you impressive insights and invigorating your work. But a few thousand words in, it starts forgetting what you told it up top, contradicting itself, getting a little lazy.

Every model has a working memory, called the context window, and it’s finite. A long, sprawling conversation or a giant pasted document fills it up, and once the model is juggling more than it can hold cleanly, the quality slips and it loses the thread (it’s the same reason a good meeting goes sideways once too many people are in the room). That same pile of context is also what you’re paying for - the bigger your context, the more expensive each chat or task becomes. More in the window means slower, pricier responses, and it pushes you toward your ceiling faster.

There are lots of little ways you can start fixing this:

• Start a fresh chat for a new task instead of letting one thread run all day, because a clean window beats a bloated one (I’ve made this mistake plenty, so learn from me).
• Feed it the slice that matters rather than the whole 90-page contract; if you only need the renewal terms, paste the renewal terms.
• When a long chat starts slipping, ask it to summarize where you’ve landed, then drop that summary into a fresh chat and keep moving, so you carry the conclusions and leave the dead weight behind.
• And match the effort to the job, since the deep reasoning modes cost more time and more tokens. Save those for the hard problems that deserve it and let the fast modes handle a quick rewrite.

One caveat that’s frustrating to hear but still matters: the providers change how limits and pricing work all the time, weekly caps one month and credits the next. Don’t get attached to a specific number, because the habits above hold true no matter how they decide to bill you.Keep your context lean and you’ll get better answers, smaller bills, and fewer walls, no matter how they reprice it next quarter.

THREE THINGS THAT MOVED (AND WHETHER YOU SHOULD CARE)

You don’t need to track all of this. So here’s the two-week version, with a verdict on each so you can skip the ones that don’t apply to you.

Claude becomes a Slack teammate. Anthropic launched Claude Tag this week, which drops one shared Claude into a Slack channel that your whole team can tag, hand work to, and pick up from wherever someone left off. It builds context from the channel over time and can keep working in the background.

• The catch: it’s on Team and Enterprise plans only, still in beta, and it replaces the old Claude in Slack app, so you’ve got until August 3 to migrate.
• The verdict: if your team lives in Slack and you’re on Team or Enterprise, pilot it in one channel now (beta means rough edges, so don’t bet the quarter on it yet). If you’re not, treat it as a clear preview of where team AI is going.

Codex now runs from your phone, on Windows too. Yesterday OpenAI made Codex Remote generally available, and it now drives Windows machines, not just Macs. Someone on your team can kick off a coding or automation task at their desk, steer it from their phone, and get pinged only when it needs a decision, so nobody’s stuck watching a progress bar. The update also leans on goal mode, where you set the outcome and let Codex work toward it, the same trick that clawed back $150 from my internet provider last issue.

• The catch: this is still a builder’s tool, most useful for someone who is beyond the beginner stage of AI usage. It’s not just for coding (despite the name), but it doesn’t require a little more experience.
• The verdict: if you’ve got a technical-curious teammate, the phone-to-desktop loop just stopped being Mac-only, which knocks down a real barrier for Windows shops.

Your AI instructions are starting to travel. Google added Skills to Gemini Enterprise, which let you package a repeatable task, like reviewing a contract a certain way or writing in your brand voice, so the AI runs it on demand. The interesting part is underneath: Gemini’s Skills use the same open format Anthropic started (a plain file called skill.md, which your technical folks may recognize). So the instructions you write to teach one AI a task are beginning to carry over to others.

• The catch: it’s in limited preview, not something to deploy this week. But if you’re running on Gemini it’s a welcome sign.
• The verdict: file it under good news if you’ve ever worried about getting locked into one vendor, because the work you put into making AI useful is becoming more portable, which lowers the risk of investing in it now.

HOW TO KEEP UP WITHOUT DROWNING

The honest problem with everything above is simple: keeping up is a job, and most leaders already have one. You shouldn’t have to read six newsletters and sit through forty demos to figure out which three things matter. That’s exactly why we built Trailblaze Academy, a growing library of short, plain-English videos that teach your team the fundamentals and the tools without the firehose.

A few that pair well with this issue:

Select Your Model: Stop Letting ChatGPT Decide How Hard to Think. The sequel to the token section above, on choosing the right reasoning depth so you’re not overpaying for a simple task or underpowering a hard one.

Custom Instructions: Stop Starting From Scratch. The fastest way to get more out of a tool you already pay for, by teaching it who you are once instead of re-explaining yourself every session.

Tiers of AI: Where Are You on the Ladder? A quick gut check on whether you’re using AI as a chatbot, a thinker, an agent, or something past that, and what the next rung looks like.

If you’re getting this email and want in, just reply with the word Academy and we’ll get you access. We’re filming new ones constantly, so think of it as a standing invitation, not a one-time offer.

THE BUZZER

The teams pulling ahead right now aren’t the ones who read every release note. They’re the ones who got a little sharper with the tools already on their desk and made learning a weekly habit instead of a fire drill. You don’t have to catch everything, just pick one move from the token section, or one video from the Academy, and run it this weekend. The rest of the noise can wait.

Thanks for huddling with us before the weekend.

Bryce Stuckenschneider

Founder, Trailblaze Labs

ABOUT TRAILBLAZE LABS

We started Trailblaze Labs because we saw a gap forming between what AI could do and what most businesses knew how to do with it. Turns out, that gap is where we live now. We help business leaders set real AI strategy, train their teams to use these tools with confidence, and build the workflows that actually move the needle. Trailblaze translates AI to the language of your business.

If something in this newsletter sparked a question, gave you an idea, or made you want to push back on something I said, I want to hear about it. Seriously. Reply to this email, connect with me on LinkedIn, or just say hi. I read everything and I love a good friendly argument about where this is all headed.

Let's keep building.