Augment Code

Introducing Cosmos: the Unified Agents Platform for software teams. Orchestrate a fleet of agents across your entire software development lifecycle, as a single organizational system instead of disconnected workers. It's changed how our own engineering team works, with throughput up 3x.

On July 22nd I am hosting a webinar on automated incident mgmt with AI, together with the folks at @augmentcode! That’s because I write a lot about my AI workflows with @tolariamd, but many people (rightfully) tell me that things are more complex in a real team. So I invited the team at Augment to share how they transformed their own engineering org, redesigning their main dev workflows. If you are interested you can RSVP here! streamyard.com/watch/CXnfrUnx…

Tune in here: youtu.be/xm8E0Kh_6o0

What does the role of an engineer look like today? In the latest episode of We Built What? we take a deep dive on the skills required. Full episode in the thread 🎧

If you do not see GPT 5.6 in your model picker, your Augment Code admin must first opt into the Responsible ZDR option in the admin dashboard, which is needed for models requiring AI Safety Programs.

GPT-5.6 Sol, Terra, and Luna are now available in the model picker! Try them today in Cosmos, our unified agent platform!

Watch the full episode here: youtu.be/5X5HclHp-Kw

The IDE is dead. Now what?

Your organization's token spend is off the charts because you've thrown everyone into the deep end and said "swim!". Of course there's deeply inefficient use of tokens, everyone is just doing their own thing. They're running loops hoping for some outcome, running local code review on every change, have agents polling for PR comments on their laptops. Not to mention the sheer volume of duplicative context, internal skills, and other support systems they've burned tokens creating. And this doesn't even account for the skill gap of effective users versus novice users. It's like the industry forgot 40 years of lessons learned on building efficient engineering teams and practices. Software engineering is a system, treat it like one and you might find tokens are actually worth the spend.

For details on how we’ve been using Fable, check out our earlier post x.com/augmentcode/st…

Claude Fable 5 has returned to our model picker, following Anthropic’s announcement. At roughly 2x the cost of Opus 4.8, Fable 5 is intended for long, multi-step tasks that require deep reasoning. In the near term, you may notice that some routine software development queries will fall back to Opus 4.8, as Anthropic tune their classifiers which block cybersecurity tasks. x.com/augmentcode/st…

Fable is back! Try it in Cosmos today.

Try it in Cosmos today: app.augmentcode.com

Sonnet 5 is now available in the model picker in Cosmos. New model, same story: better performance for less money. Sonnet 5 begins to close the gap to frontier-level intelligence, bringing near-Opus performance at Sonnet pricing. This means that quality and cost become less of a trade-off, and we recommend trying Sonnet 5 for work where cost and throughput matter most. Until August 31, experience this new level of Sonnet intelligence at a discount ($2/M input tokens, $10/M output). After this period, Sonnet 5 will be priced the same as previous Sonnet models.

@GergelyOrosz @karpathy This is internal feedback on our version (Cosmos). Cloud agents + integrations + Slack.

Full conversation here: youtube.com/watch?v=0fX03K…

How do you keep up with AI? @gregorojstersek has advice for any engineering leader:

You can opt in to receive the benchmark when it is published.

The next wave of AI adoption is a measurement problem. AI tooling is now a line item in every engineering budget. But, assessing its value still remains ambiguous. Most engineering leaders went into AI expecting 2-3x productivity gains, but are landing closer to 30%. We are building a benchmark on AI engineering productivity and ROI. If you help lead, measure, or operate engineering work, your input helps set the baseline. Take the survey here: augmentcode.com/lp/ai-engineer…

The human reviews the design before any code exists. Our Code Review experts then read the full diff in minutes and check it against that approved design. Implementation size stopped being a constraint the moment a person was no longer the one reading every line. The unit of work is no longer the PR. It's the project, and one agent can now carry a project from approved design to merged code on its own. Full write-up: augmentcode.com/blog/accelerat…

A few we shipped to production like this, all on systems with thousands of active users, none of which caused an incident.