More agents ≠ engineering transformation. Meet Cosmos, the OS for agentic software development.
- Tweets 1,765
- Following 55
- Followers 22,362
- Likes 2,438
Your PR sat for three days. CI was red, comments piled up — and your reviewer didn't have the confidence to hit approve.
A fleet of agents performs multiple rounds of review, fixes the CI, the comments, the conflicts, then hands your reviewer a deep review briefing and end-to-end evidence.
Humans still approve and merge 👇
10
6
21
2,735
Quick start: a software factory in Cosmos is two screens and one prompt.
1. Connect your repo
2. Create an environment
3. Describe the factory you want
Advisor picks the experts, writes the configs, applies them, and asks you when it needs a decision.
1
1
19
3,040
The best part of an agent that runs in the cloud isn't the speed. It's that you can walk away.
Cosmos notifies you the moment it needs you. Everything else, it handles.
2
4
26
3,077
Watch a PM and an engineer collaborate on a PRD — with their agents, without leaving Cosmos.
A PM drafts it with her agent. An engineer opens the same file, has his own agent review it, leaves a comment. Her agent answers it and updates the doc.
One filesystem. Both people. Both agents.
4
6
21
2,286
GPT-6 Astra is live in Cosmos!
OpenAI's most capable model yet brings a significant jump on long-horizon coding and multi-step agentic work.
Give it your hardest task to work on over the holiday weekend at augmentcode.com!
2
8
11
1,984
Most agent platforms make you learn the platform.
Cosmos Advisor just answers. Build this. Fix that. What did it cost.
One prompt box, your entire software factory. Try it today ↓
1
7
18
2,355
Claude Fable 5.1 is now available in our model picker! We're starting off with the same use cases as Fable 5 - extremely long, multi-step tasks that require deep reasoning - and we're increasingly pushing the model to complete tasks that can run unattended.
Try it today in Cosmos: augmentcode.com
1
1
14
1,834
Opus 5 is live in Augment Code. In our internal evals it scored higher than Opus 4.8 on correctness, completeness, and code reuse.
We trust it enough that we're upgrading one of our Cosmos code review experts that reviews all our code to Opus 5: near-Fable review quality at a lower price.
Try it in Cosmos today!
Introducing Claude Opus 5.
It's a thoughtful and proactive model that comes close to the frontier intelligence of Fable 5 at half the price.
1
4
19
3,698
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 🎧
1
1
6
1,937
How do you keep up with AI? @gregorojstersek has advice for any engineering leader:
2
4
13
1,998
A few we shipped to production like this, all on systems with thousands of active users, none of which caused an incident.
1
2
1,076
Most teams point agents at small, well-scoped work. A ticket goes in, a tidy PR comes out, a human reviews it, it merges.
Our largest projects don't work that way.
On Cosmos, once a human has reviewed and approved the design doc, we hand the entire design to a single worker agent and let it build the whole thing as one PR. It implements the design, fixes its own CI failures, and works through review comments until the change is ready to merge. No ticket-splitting, no relay between agents.
The PRs are large, often thousands of lines, and we prefer them that way. One agent holding the full approved design writes a more coherent change than several agents each working from a slice of the context. The seams you get from splitting, the mismatched assumptions between unit three and unit seven, don't show up when one agent carries the whole picture.
4
4
33
2,720
"There's never been a better time to be a software engineer", @JustinReock with the hot takes 🔥
Watch the full episode here: piped.video/n-VGqrRK2vk?si=m-gR…
3
7
1,111
TL;DR: Fable 5 isn’t the right-sized model for every task, but when quality and depth matter (review, architecture, long-running changes), the lift is impressive.
Here’s how it stacks up vs Opus 4.8 + GPT‑5.5
2
2
4
1,108
Fable 5 is now available to Augment Code users. At ~2x the cost of Opus 4.7, Fable 5 is a premium option for long, multi-step engineering work.
Try Fable 5 today in Cosmos, our unified agent platform, to fully experience the power of the most advanced model on the market.
5
6
34
5,378
We're all experiencing the moving bar of "being AI native". How do we navgiate it?
Watch the full episode with @CircleCI's @z00b here: piped.video/2cGdwowJiSw?si=WRTk…
2
6
760
Live this Friday, June 5 at 10 AM PT: the first look at Cosmos, our new Unified Agents Platform.
Your engineers are shipping more code than ever, but your org doesn't feel 10x more productive. Adding more agents won't fix that. You need a system.
Join @VinayPerneti (VP of Eng), Rich Hankins (Founding Engineer), and Sharath Rao (Solutions Architect) to see what changes when agents share context and memory across the team.
Sign up to join live or get the recording: watch.getcontrast.io/registe…
2
3
10
3,131
See it live this Friday, June 5 at 10am PT. Full walkthrough of Cosmos, plus how our own team at Augment uses it.
Try Cosmos today: cosmos.augmentcode.com
Register here: watch.getcontrast.io/registe…
1
4
757
Cosmos runs in your environment or ours, supports the models you choose, and provides the observability, auditability, and human oversight required to deploy agents at scale.
Learn more: augmentcode.com/blog/cosmos-…
1
2
844
At the core is Augment's Context Engine. It helps agents understand how a codebase actually works, not just which files match a keyword search.
Same model, half the token bill. Better context, better decisions, higher-quality work.
1
1
346
Cosmos gives your organization a single platform for running agents.
Deploy our tuned agent experts for triage, implementation, review, testing, and incident response, or build your own. They share context and memory, coordinate with one another, and bring in humans where judgment matters.
1
1
4
1,097
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.
10
14
59
10,253
Opus 4.8 is now available in Cosmos.
In our evals, the model demonstrates strong performance on long-running tasks - including multi-hour executions and ticket-to-PR workflows with minimal intervention.
Use Cosmos and Opus 4.8 on your most complex @Linear tickets or challenging @getsentry issues!
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.
Available today at the same price.
ALT Benchmark table showing how Claude Opus 4.8 compares to its predecessor and to other models on tests of coding, agentic skills, reasoning, and practical knowledge work tasks.
6
5
25
5,010
The team building Codex at @OpenAI has a front row seat to how engineering organizations are changing, as agents become part of the team.
Join a live fireside chat on Thursday, May 21 at 10 AM PT with @TheRohanVarma, Product Lead for Codex, OpenAI, and Vinay Perneti, VP of Engineering, Augment Code, as we discuss:
· How the Codex team builds software end to end with agents
· What's actually shifting inside engineering orgs adopting them
· Which change management tactics are working, and which aren't
If you lead an engineering team and are trying to make sense of this moment, this is the conversation for it.
Register today:
watch.getcontrast.io/registe…
2
12
2,254
Uber's internal agent platform now generates 11% of all PRs across the company.
Join a live conversation on Friday, 5/15 at 10 AM PT with Nikhil Ramakrishnan, Senior Software Engineer at Uber. In our discussion, he will uncover the platform foundation and reveal how engineering work at Uber has actually changed.
See how Uber built their agentic developer platform, and how engineers actually use it.
Register today:
watch.getcontrast.io/registe…
1
12
1,931
"Excited, anxious, invigorated."
That's how one engineering leader described going AI-native. We asked 218 others how they feel.
The results? Remarkably consistent:
"Insecurity, mistrust, hope."
"Optimistic, excited, threatened."
It's the same people, holding many feelings at once.
That's the baseline running through our new report, The State of AI-Native Engineering in 2026, co-authored with @VinayPerneti and @EmmaStarks.
1
8
22
3,074
We surveyed 219 engineering leaders in April 2026 and asked them what's actually happening inside their organizations.
The finding that stuck: most have adopted AI tools. Most haven't changed how they build software.
That gap, between tool adoption and actual transformation, is where most engineering orgs are sitting right now. And the leaders who've crossed it have a few things in common.
We're hosting a webinar to walk through the data. What it shows, what surprised us, and what it suggests you should do next.
Register here for our session on May 14 at 10 AM PT:
watch.getcontrast.io/registe…
1
3
16
2,124
Harvey built their coding agents to be collaborative by design. Engineers, PMs, and researchers all contribute context to the same agent, not parallel ones running on separate laptops.
They published a blog post on how it works. We're going behind it.
We're hosting a live conversation with Joey Wang, Engineering Lead at Harvey, to dig into what the blog post didn't cover: the architecture decisions, the tradeoffs, and what didn't work the first time.
If you're thinking about where cloud agents fit into your engineering org, this is the conversation to be in.
Register here for our session on May 5 at 10 AM PT:
watch.getcontrast.io/registe…
1
1
5
1,672
Building a model router is non-trivial. The hard part isn't picking - it's switching.
Prism's job is to switch only when the expected win from a different model exceeds the cost of the cache eviction.
1
1
353
We know that developers and teams have strong preferences for different model families: with Prism, you can stay in the model family you like, at lower cost.
→ Prism (GPT + Kimi) targets GPT 5.5
→ Prism (Claude + Gemini) targets Opus 4.7
2
2
2
743
Today we're shipping Prism: a new option in the Augment model picker that efficiently routes each turn to the model that fits the work.
On our internal multi-turn coding benchmark, Prism matches the best individual model on quality at 20–30% lower cost per task than frontier models.
3
6
55
5,087
Every runner used fewer tool calls and finished faster. The agents found what they needed in fewer lookups.
Output tokens fell by similar margins across runners. Per PR, Karpathy was faster and cheaper on about 30 of 40 PRs. The pattern held across all three agents.
A 3–10% efficiency gain from a small prompt change isn't a model breakthrough, but if you're running a coding agent at scale, it's real money, real latency, and real capacity.
458
Quality: basically unchanged for Auggie and Codex. Claude Code dropped −0.07, with more conservative trajectories and ~5% fewer files touched per task.
Karpathy-style guidelines don’t transfer uniformly across agent harnesses and repositories.
In Codex, the guidelines likely add useful structure (improving efficiency).
In Augment, the baseline prompt already encodes similar constraints, so the marginal impact is smaller.
In Claude Code, the system prompt may already be highly constrained, so layering additional constraints could reduce exploration and degrade performance.
2
2
3
1,546
We added @karpathy -inspired coding rules from @jiayuan_jy to AGENTS.md and ran 40 @openclaw PRs through three coding agents.
The result:
Code quality was basically unchanged, but the agents got there with less work. Fewer tool calls, lower time and cost.
4
14
143
16,463
Most engineering orgs have adopted AI coding tools.
Far fewer have changed how they build software.
There's a difference between adding AI to your workflow and rebuilding the workflow around AI. We're hosting a session that discusses an engineering team that transformed their SDLC, and sharing exactly what it took.
If you're leading an engineering org and trying to move from experiment to operating model change, this one's worth your time.
Register here for our session on May 1 at 10:30 AM PT:
watch.getcontrast.io/registe…
1
3
18
1,770
The biggest mistake engineering leaders are making with AI is treating codegen as the whole transformation.
Engineers only spend ~16% of their time actually writing code.
So even "perfect" AI codegen only attacks 16% of the system.
The real leverage is in context, workflows, review, docs, architecture, and removing bottlenecks.
1
4
19
2,573
GPT 5.5 is now available in Augment Code!
It has topped our internal benchmarks.
Faster. More reliable. Better outputs.
Try it now!
8
5
120
7,032
Opus 4.7 is now the default model in Augment, and it’s 50% off until April 30!
Async workflows and agent orchestration are becoming the real bottleneck in AI-powered development.
Models can generate code quickly. But long-running tasks drift, CI/CD breaks in non-obvious ways, and multi-agent work falls apart halfway through.
Opus 4.7 is the first model we’ve used that feels built for this.
2
7
35
4,513
We're excited to share that Intent, our newest multi-agent orchestration platform, is live on Product Hunt today. If you’ve been enjoying what we’re building, it would mean a lot if you could support us🥳
Check us out here 👉 producthunt.com/products/aug…
12
5
61
5,835
“How can I lead my engineering team to increase output without losing velocity?”
That’s the question we’ll tackle in our next live webinar on Thursday, April 9, 2026 at 10:00 AM PT.
We’ll cover:
· Where AI actually moves the needle across the SDLC
· How to use AI to cut toil while keeping quality and ownership high
· The operating models top teams use to scale AI sustainably
If you’re leading an engineering team, you don’t want to miss this.
Grab your spot today: watch.getcontrast.io/registe…
1
8
1,405
Join our livestream tomorrow with @DynamicWebPaige from Google DeepMind where we put Gemini 3.1 Pro to the test and show you what it's capable of.
Sign up to join live or receive the link to the recording.
Last chance to register: luma.com/vfpdjkym
1
6
3,289
Gemini 3.1 Pro is now available in our model picker.
Its performance is comparable to Opus 4.6 on real-world engineering tasks, at 2.6x lower cost per message.
We’re especially excited about how it performs on:
- Planning and reasoning through changes across large codebases
- Debugging and investigation
- Navigating unfamiliar systems
For planning and investigative work, try Gemini as your first pass.
2
7
65
5,358
So we rebuilt code review for how engineers actually behave in 2026.
Introducing SwipeReview™ by Augment Code
- Short video PR summaries
- Swipe right = approve
- Swipe left = request changes
1
1
1
510
