Hi! I'm the guy putting ads in your editor (VS Code and CLI) with Amp. AKA, the worst person in dev tools. I thought, amidst all the Anthropic / ToS drama and Amp launching a new and updated free offering, I'd take the opportunity to talk about our ads network. (Long post; sorry in advance) A little about the free offering: > you get $10 in free credits every day (up to $300 / month! | $3,600 / year!) > use them however you want with Smart mode (Opus 4.5), Rush mode (Haiku), or specialized subagents like the oracle (GPT 5.2) and the librarian (Sonnet 4.5). > we replenish credits at $0.42 per hour > If you run out, wait an hour or continue with pay-as-you-go How can we do this? Simple: Enterprise customers and **ads**. These fund free usage for users, teams, and even enterprises. More about our ads network: > we only partner with reputable dev tool companies, so you aren’t going to get a Burger King ad while debugging a YAML file (see screenshot below). > ads are unobtrusive by design so they don't f&*k with your focus and flow > your code is safe; our ads network follows all of our established security conventions (link in comments) > we don't train on your data to service ads > if you don't want ads (or, free credits) just pay for usage Why did we go this route? > inference. is. expensive. > if it isn't painfully obvious, a $200/month, or $20/month plan doesn't pay for the amount of tokens people actually use with these tools > this is why we viewed the AI coding space as a house of cards: every company (ours included) was and is subsidizing usage to attract users. > and, it works / worked! Cursor, Claude Code, Cody, etc. all did it, but it felt unsustainable > so, we experimented with product changes, pricing & packaging changes, and business model changes to find something sustainable that wouldn't leave us beholden to another company's whims > after many failed experiments, we realized subsidizing inference with ads $$$ (not VC $$$) was the best, most sustainable way to provide an accessible and high quality product to as many users as possible > quick aside: it's funny how we got here... it was a running joke internally that ads would solve all our problems, hell, one of our competitors made an entire april fools joke about burger king ads in the editor > despite this... slowly... and then suddenly, we realized ads **were** the best path forward, but only if done 'right' > so, we built our ads network on the principles that our ads must be: unobtrusive, highly relevant, and immediately useful Where are we today? > we serve ~10 million ad impressions monthly > we have received (and I'm keeping an actual log of this) **7** complaints about ad quality and disruptiveness since our launch (internally... it's much higher, but that's how we ensure a high quality bar) > some complain about the exchange of free inference for ads, but you can always pay instead! I think we've threaded the needle well, but we are constantly iterating. If you are interested in becoming an ads partner to help drive this crazy experiment and put your tools and services directly in our users editors, click the 'become and ad partner link' in the comments. Lastly, a MASSIVE thank you to all of our advertising partners who make this crazy experiment possible. My DMs are wide open. Blow me up on the timeline. Whatever works best for you, just please share any and all feedback you may have!

I used AI to explain the Anthropic drama to my girlfriend, with fruit.

OK, killing this. I hate the idea of an iron curtain on the frontier. You all do, too. I don’t want to normalize it. I'm sorry and appreciate the response. We still must follow gov't and model lab rules, but each new rule and distraction makes the open frontier more appealing.

The new beta Claude Code model switcher

Definitely

WOW! This has completely changed my perception of the model. This experience affirms to me how important the harness is for these models. It's probably why @AmpCode is so awesome. They take the raw model inference and tune it just so.

It's funny that OpenAI might align models better by focusing less on alignment and more on precise instruction following.

Do the Kroenkes understand that Altitude salaries don't count against the salary cap?

My 2026 Nuggets offseason preview is here! 70 pages of content this year, covering: -10 Mock Offseasons -Projecting Peyton Watson’s deal -Will Denver duck the luxury tax? And more! Link to read / download: drive.google.com/file/d/1DZX…

Claude Fable 5 in Amp (via a plugin, built-in soon)

amp update amp plugins add --auto-update @amp/fable-mode amp --mode claude-fable-5

Now that, that's an IDE

TTFT down 87%, 32% faster responses, end-to-end task completion up to 40% faster

Hell, it's from the Inner Richmond to Bernal.

Protein tortilla with sliced chicken, buffalo sauce, and some cheese.

this guy gets it

How do we compare model perf in ARC-AGI-3? In most benchmarks you just compare scores, but with ARC-AGI-3 you get reasoning logs across all the games you play To compare Opus 4.8 to Opus 4.7 we used LLM as a judge Using @AmpCode (my daily driver right now) I set up a skill to compare models, then it spawned a sub-agent per game per model Each sub agent did a single-game analysis, then brought its notes back to the main agent Very cool to see all of this come together. It would have taken 2-3 days of analysis by hand before

Genetics consistent calories?

I love when technological progress kills a feature. The models are so good that (A) this feature is too risky and (B) it's not needed anymore to convince people that agents are good. ampcode.com/news/end-of-publ…

My wife and I recently celebrated 2 years of her being cancer-free. The New York Post wrote about her experience with @Biograph, and I think it’s worth the read: nypost.com/2026/06/01/health… My take: About 2 years ago, we decided to try Biograph. It’s expensive, and we had real guilt about how inaccessible services like this are for most people. We also had concerns about false positives, over-testing, and the anxiety that can come with finding things that may or may not matter. But we are both pretty health-minded, and as we’ve moved firmly into “middle age,” we thought it would be useful to get a deeper look at our health -- and probably kind of fun! Fun is not how I would describe our experience. Informative, scary, effective, and life-changing is closer as one of the tests found a very early, stage 1 tumor on her pancreas and spleen. It was an incredibly scary and tense period for our family, filled with a lot of “what ifs.” What if we hadn’t done the screening? What if we had found it later? What if the treatment path had been much harder? Because it was caught so early, and because of the incredible care she received through UCSF and Stanfurd (go Bears!) Medical, surgery was enough to treat it without chemo and radiation. We've spent a lot of time talking about the impact Biograph had on our family. It was an incredibly messy and stressful time of life. Things worked out for us, but it was still really, really hard. We ultimately landed on this: Biograph is a provider of information about your body and your health. Some of that information is useful. Some of it isn't. Some of it can send you down a road to nowhere, while others can change your life for the better. For us, more information was better. It gave us the ability to make informed decisions and take action before things got much worse. If you are still reading, I want to be clear that I'm trying hard not to make this sound like an ad (it isn't!) so I want to speak to some of the downsides. These services are expensive. They are out of reach for the vast majority of people. That really sucks. It isn't fair. And, it is yet another example of how broken and unequal our healthcare system can be. Full-body scans can also find things that may never cause harm, which can lead to anxiety, follow-up testing, unnecessary procedures, and more cost in an already strained healthcare system. They are also not for everyone. If you are prone to fixation, have a hard time sitting with uncertainty, or know that ambiguous medical findings would be emotionally overwhelming, it is worth thinking carefully before doing something like this. For us, though, @Biograph likely saved my wife’s life. Or, at minimum, saved her from a much worse and more complicated outcome. And for that, I’m incredibly grateful.

one of the best @AmpCode feature