AI has entered its allocation era: Companies are starting to ask who gets access to the most powerful models, what those models should be used for, and when the cost is worth it. Today, head of tech consulting Mike Taylor argues that token budgets may start to look like trading portfolios, with the biggest compute budgets going to the people who can prove the biggest returns. Elsewhere, head of operations Arielle Shipper stress-tests her own AI habits, head of growth Austin Tedesco shares how he writes with Spiral and Codex, and senior designer Daniel Rodrigues tracks the resurgence of trad design on social media.
Was this newsletter forwarded to you? Sign up to get it in your inbox.
Signal
ROI comes for tokens
Just a few months ago, heavily subsidized AI plans and a manic push by corporations to get their employees familiar with the technology led to aggressive tokenmaxxing, or measuring AI adoption by how many tokens employees use.
That era appears to be over. Shifting pricing models from the frontier labs, the transition from chatbots to long-running agents, and the emergence of powerful, incredibly costly new models have converged to create massive AI enterprise bills, often without tangible results. Uber, Meta, Amazon, and Walmart have all moved to place caps on employee AI use.
These companies will still spend big on AI. But they’re beginning to think more
strategically—or restrictively— about how to get capable, token-hungry models and workflows in the hands of people who can maximize the return on investment (ROI).
One solution is to give engineers a set percentage of their salary to spend on tokens per month. Head of tech consulting Mike Taylor thinks the financial industry offers a possible model: Capable engineers will receive multiples of their salary to manage on token spend just as financial traders manage portfolios many times the size of their annual compensation. Getting the most out of current models already requires real capital—recently, Cora general manager Kieran Klaassen spent $2,000 in Cursor credits in the span of a couple of days, a number that will look quaint as the models continue to improve.
Mike predicts that tokens will be allocated and controlled on a stricter basis under this system: “Just like with trading, you’ll have risk limits, auditing, and you’ll have to get certain-size bets approved.” More broadly, we’re likely headed for a world with more restrictions placed on who is granted a token budget and what they’re allowed to do with it. Top engineers who can prove ROI will benefit, as will managers. “Who are you going to give a token budget to, an IC [individual contributor] who prefers to handcraft their code anyway, or a manager who is already accustomed to directing 5x or 10x their own salary in resources?” Mike asks.
In this version of the near future, an intern might have access to Composer 2.5, most employees will work out of Codex or an equivalent, while Fable-grade models are reserved for top engineers. There’s too much money on the line to hand over frontier access to anyone but an elite few.
Attio is the CRM for teams who set the pace
It compounds every customer signal into context, then acts across your funnel to let you move at unmatched speed and scale. With agents and automations that build pipeline, chase signals, and move deals forward, Attio orchestrates your revenue work around the clock.
Loved by high-growth startups like Granola, Modal, and Wispr Flow, Attio amplifies what you can achieve.
Discuss
“Can we downgrade the model a little bit and still get the same outcome?”—Joel Neeb, chief transformation officer of the software company 8x8, in Wired
Even aggressively AI-forward companies like 8x8 are testing ways to rein in token spend. Wired reports that 8x8 encourages all 1,800 full-time employees to use Claude and tracks usage on internal dashboards. But as Opus 4.8 drives up spend—it costs 1.7 times more than a model Anthropic released earlier this year—the company has discussed requiring employees to prove older models cannot do the job before they get access to the newest one.
Data point
6.6
That’s head of operations Arielle Shipper’s current level on Every’s eight levels of AI adoption, according to a weekly Codex review she set up for herself. Her prior baseline was 5.5, or agent-first but not orchestrating multiple agents or workstreams at once.
Arielle uses Codex to manage work that used to require a lot of manual context switching, everything from weekly-sync updates to credit card provisioning in Ramp. After asking Codex to assess her AI usage, she turned the recommendations into a recurring Friday automation. It reviews her sessions from the past week, tells her where she is on the ladder, and gives her concrete tactics to try next. One tactic that stuck: Arielle used /LFG, compound engineering’s agent workflow for taking a goal through planning, execution, review, and improvement, to map out her weekly ops update system.
Try it yourself. Paste the below prompt into Codex:
Then ask:
Steal this workflow
Writing with AI
For certain types of writing, like an end-of-Q2 sprint strategy document, head of growth Austin Tedesco is happy to have a well-harnessed agent generate as much of the text as possible.
For more personal projects, like his weekly Substack food newsletter, his process is different. He’s set up OpenClaw and Codex to serve as a thought partner and editor who helps him store, organize, and sharpen his thinking. Here’s his AI writing workflow:
- Create a standing writing file. Austin’s lives in Proof, Every’s agent-native document editor, and has three sections: “Ideas bank,” for rough thoughts, “Outline,” where ideas start to take a formalized shape, and “Draft,” where an outline is fleshed out into a draft essay.
- Collect half-formed thoughts. When an idea strikes, Austin sends it to his OpenClaw via text, even if it’s messy or half-baked. Because the agent is connected to Every’s AI-writing assistant Spiral, which is trained on his tone and style, it can refine rough ideas into prose that matches his writing style. He’s also hooked it up to his Substack archive, so it can pull relevant context from previous editions, like when he’s already mentioned a restaurant or if his position on a particular topic has changed.
- Knock out a draft. By the time Austin opens his laptop, his writing file already contains ideas he’s had throughout the week, organized into a detailed outline. He then opens up the Proof doc in Codex’s in-app browser with enough context to knock out a publishable post in one sitting.
- Monologue your way through writer’s block. While Codex handles the idea collection and helps with context and outlines, Austin mostly writes the prose himself. If he hits a wall, however, he’ll open Monologue, Every’s voice dictation app, and brain dump his way through. In these voice notes, he often acts as an editor, articulating where and how the draft doesn’t align with what he’s trying to articulate. From there, he has Codex take his thoughts, clean them up, add transitional phrasing, and incorporate them into the piece.
“A lot of writers would say, ‘That’s AI bullshit,’” he says. “But those are my words. Monologueing is great for breaking writer’s block, and even better for crafting an outline.”
Try it this week:
- Pick one recurring writing project and create a doc with three sections: Ideas bank, Outline, and Draft.
- Give the model enough context to understand how you write. Use Spiral to generate a style guide from previous pieces, or paste a handful of representative samples into Codex or Claude and ask: “Create a short style guide from these examples. Capture my tone, structure, sentence length, recurring moves, and things I avoid.”
- Give it a small archive: links, excerpts, or titles from past pieces that might be relevant. Follow up with: “When I send new half-formed ideas, flag where they connect to something I’ve written before, where my thinking has changed, and where I might be repeating myself.”
- Start feeding it messy notes, links, and voice transcripts. Ask it to sort them into the three sections, update the outline, and use the material to either write a draft yourself or ask the model for a first pass.
Inside Every
Handcrafted versus agent-generated
Senior designer Daniel Rodrigues has noticed a bifurcation in the design content served to him on Instagram. Some posts display designs made with an elaborate array of AI tools and workflows, and a separate vein of content spotlights painted, illustrated, or other hand-created visuals.
Creators and consumers of online content tend to pick a lane: They either want content that highlights cutting-edge AI design or they have an aversion to all things AI-generated.
It’s not just design. My LinkedIn feed feels like a tug-of-war between AI-pilled writers and those who vehemently post about never using AI—to varying degrees of plausibility. Anecdotally, writing of all flavors increasingly feels either written by Claude or deliberately weird and personal. A similar dynamic is at play within engineering, which encompasses agent-orchestration workflows and a return to “hand-crafted” code.
Daniel is drawn to AI maximalism and physically-made content. AI tools like Midjourney, Unicorn Studio, and Claude Code have greatly expanded his abilities. “I feel more powerful than ever,” he says. But when people compliment the framed images on his wall during Zoom calls, many of which he created with AI help, he feels strange taking full credit. The work feels more akin to curation than creation—hence the desire to pick up a brush and create something he is solely responsible for.
Laura Entis is a staff writer at Every. You can follow her on LinkedIn. To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.
For sponsorship opportunities, reach out to sponsorships@every.to.
Help us scale the only subscription you need to stay at the edge of AI. Explore open roles at Every.
