wiynt

saving bookmarks

[Praveen Neppalli @praveenTweets]

Agentic AI adoption is on fire at @Uber, and it's changing the way we build, not just in engineering, but across the entire company.

Today, 99% of our engineers use AI tools. More than 70% of pull requests are attributed to local or cloud agents. And our engineers have built 2,500+ agent skills across the software development lifecycle.
Those numbers are exciting, but they led us to a much bigger question:

How do we bring agentic AI beyond engineering?

Finance. Legal. Operations. Marketing. Customer Support. HR. Procurement.

These functions run on complex workflows that are often manual, highly nuanced, and spread across dozens of systems. You can't automate them effectively by looking at process diagrams or documentation. You have to understand how the work actually gets done.

So we created something called Agentic Pods.

The idea is simple.

We handpicked ~30 of our most AI-proficient engineers (people with deep knowledge of Uber's systems) and paired each of them with a domain expert from a business function.

Then we gave every pod just two weeks.
• Days 1 – 2: Shadow the expert. Observe every step. Document workflows. Ask questions. Build intuition.
• Day 3: Prioritize opportunities based on scale, repetition, business impact, and data availability.
• Days 4 – 5: Build a working agent alongside the person doing the job.
• Days 6 – 9: Validate with several others performing the same work. Does it generalize? Does it actually make their job better?
• Day 10: Ship.

In just the past two months, we've run 16 Agentic Pods across 16 different business functions.
• Capital allocation across 150 cities: 15 hours → 30 minutes.
• Financial pacing reports: 2 days → 10 minutes.
• Marketing web quality assurance: 2 weeks → 50 minutes.
• Support workflow creation: 9,000 manual workflows → self-service automation.

The productivity gains are impressive, but what surprised us most wasn't the speed.
• It was how quickly engineers embedded in unfamiliar domains uncovered opportunities that had been hiding in plain sight.
• The biggest wins rarely come from automating one task. They come from rethinking an entire workflow. Once you redesign the workflow around AI, you often eliminate handoffs, remove unnecessary approvals, replace legacy tooling, reduce vendor spend, and dramatically accelerate decision-making.
• The workflow becomes the unit of automation - not the individual task.
• The most impactful agent skills cut across teams, orgs, functions, tools, and systems.

The biggest lesson? The best AI opportunities are rarely visible from the outside.

You discover them by sitting next to the people doing the work, understanding every friction point, and building with them, not for them.

We're now forming a dedicated team to scale this further and go deeper. They'll deeply understand the work, redesign it from the ground up, and use AI to fundamentally change how the business operates.

It's exciting times!

[Mark Zuckerberg @finkd]

(1) Today we're releasing Muse Spark 1.1 -- a strong agentic and coding model at a very low price. It's available through our new Meta Model API and in Meta AI.

[corbin @corbin_braun]

fable 5 is making me realize that when AGI is here.

99% of humanity still won't use it or care

crazy.

[Palantir @PalantirTech]

Our expanded thoughts on an institution’s freedom to pursue new opportunities, its economic rights, and its ability to expand them in the age of AI.

Please find attached our white paper, Institutional Sovereignty in the Age of AI, outlining the 15 steps governments and companies must take to protect both their sovereignty and their alpha.

https://t.co/rZuNeOKXU9

[ClaudeDevs @ClaudeDevs]

https://t.co/v694m8Eaj6

[Shiva @_Shiva_iitp]

Who's gonna tell vibe coders about...

> Rate Limiting
> Caching
> Load Balancing
> Reverse Proxies
> API Gateways
> CI/CD
> Docker
> Kubernetes
> Service Discovery
> Circuit Breakers
> Timeouts
> Retries
> Exponential Backoff
> Idempotency
> Message Queues
> Pub/Sub
> Event-Driven Architecture
> Distributed Transactions
> Saga Pattern
> Dead Letter Queues
> Cron Jobs
> WebSockets
> Long Polling
> Server-Sent Events
> Database Indexing
> Query Optimization
> N+1 Queries
> Connection Pooling
> Read Replicas
> Sharding
> Partitioning
> Replication
> Leader Election
> CAP Theorem
> Eventual Consistency
> Optimistic Locking
> Pessimistic Locking
> Distributed Locks
> Race Conditions
> Deadlocks
> Memory Leaks
> Garbage Collection
> Thread Safety
> Backpressure
> Autoscaling
> Horizontal Scaling
> Vertical Scaling
> CDN
> Edge Caching
> Cache Invalidation
> Feature Flags
> Blue-Green Deployments
> Canary Releases
> Rolling Deployments
> Rollbacks
> Health Checks
> Liveness & Readiness Probes
> Monitoring
> Logging
> Distributed Tracing
> Metrics
> Alerting
> SLOs
> SLIs
> Error Budgets
> Observability
> Secrets Management
> IAM
> OAuth
> JWT Rotation
> TLS
> Encryption at Rest
> Encryption in Transit
> WAF
> DDoS Protection
> CORS
> CSRF
> SQL Injection
> XSS
> SSRF
> Database Migrations
> Schema Versioning
> Disaster Recovery
> Backups
> Failover
> Multi-Region Deployments
> Chaos Engineering
> Cost Optimization
> Cold Starts
> Serverless Limits
> Latency
> Throughput
> P99 Latency
> Tail Latency
> Network Partitions
> Clock Skew
> DNS
> TCP vs UDP
> HTTP/2 & HTTP/3
> gRPC
> Webhooks
> API Versioning
> Semantic Versioning
> Infrastructure as Code
> Terraform
> Helm Charts
> Build Caching
> Dependency Hell
> Production Incidents
> On-call
> Postmortems

...until the app gets its first million users.

[Zack Voell @zackvoell]

boss: "work harder"

parents: "move out"

girlfriend: “quit being a loser”

codex: "holy shit that's brilliant"

[Scottie Pippen @ScottiePippen]

Sometimes it takes 10 years to get that 1 year that will change your life. Don't give up.

[toly 🇺🇸 @toly]

Any idiot can give money away, and the same amount of problems remain as before. This is midwits celebrating mediocrity.

If $26b was invested into startups that resulted in net new 1 trillion of productive capital, roughly 1 million people would have been permanently lifted out of poverty globally.

As any Soviet child can tell you, the emancipation of the proletariat can only be achieved by a massive increase of productive forces. Any spending that doesn’t result in an economic surplus destroys the capital basis that the proletariat revolution can be built.

[OpenAI @OpenAI]

We’ve designed and built our first AI chip: Jalapeño.

Designed from the ground up by OpenAI and brought to production with @Broadcom, Jalapeño is purpose-built for the LLM workloads powering ChatGPT, Codex, the API, and future agentic products.

Chips are foundational to the AI economy. Building our own expands our full-stack platform from products to models to infrastructure, and will help us scale intelligence, serve more people, and expand access to AI.

[Marc Andreessen 🇺🇸 @pmarca]

Cyberpunk AF.

[Alex Corrino @AlexCorrino]

This is a hell of a chart from @Nomura

[Haseeb ＞|＜ @hosseeb]

https://t.co/C8s3FTMVzn

[hardmaru @hardmaru]

I’m proud to officially launch Sakana Marlin, our first commercial product: an autonomous “Ultra Deep Research” agent designed to act as a Virtual CSO.

Try Marlin: https://t.co/4PPTM8Fc2a
Blog: https://t.co/HqVMkU1Juf

Marlin isn’t just another deep research assistant. It is the direct productionization of our team’s core breakthroughs, including our AB-MCTS work (NeurIPS 2025 Spotlight) and The AI Scientist (published in Nature).

Instead of generating text in seconds, our focus is on long-horizon, sample-efficient reasoning.

Marlin scales up inference-time compute to execute up to 8 hours of continuous, autonomous reasoning. It forms hypotheses, navigates the web, resolves contradictions, and delivers exhaustive, expert-level strategy reports and structured slides.

[sui ☄️ @birdabo]

‼️BREAKING: a tiny lab in Tokyo just built a model that matches the performance of Fable and Mythos.

the numbers are real too - 54.2 on SWE-Pro, 95.1 on GPQA-D, 93.2 on LiveCodeBench v6. it edges out Opus, Gemini 3.1, and GPT 5.4 on each one and because it’s just coordinating existing models, it delivers Mythos-class capability with zero export control risk.

seems like another lab is breathing on Anthropic’s neck again.

‘Fugu Ultra’ is also available to use in codex.

Are you not entertained yet?

[goodalexander @goodalexander]

The thing people are missing on GLM 5.2

Before it would have been kind of useless to rent H200s bc the open models weren’t good and codex subsidies so high no practical need to

But subsidies getting cut. And GLM 5.2 is good enough to want *way* more tokens bc looping doesn’t break

Not to mention god knows what data you’re going to lose now the companies are retaining your data more for “national security”.

Completely changes economics in structural semi permanent way and it’s worth noting each of these points are likely to accelerate

1) looping (massive token consumption with sub agents) will get better with model capability and base error rate bc compounded error
2) coding plan subsidies will get cut as pressure builds into IPOs to show margin improvement
3) national security concerns will increase causing larger data retention windows and more IP risk and perverse incentives (Anthropic building the software you use Claude code to build)

Providers like Vast are already seeing insane hockey stick acceleration that started in Feb and is probably going to hit a parabola

This creates a flywheel for discount and efficiency providers which are highly international by nature (Malaysia bought billions of GPus, Middle East)

It’s worth repeating: prior to glm 5.2 this was all kind of theory / LARP. Just bc the open source models got brutally outperformed by close source

I’m incredibly jacked up. Rising tide lifts all boats. GPUs. Chip Marketplaces. Chip financing. Crypto mining. The future is now

[Students For Liberty @sfliberty]

In July 1985, over a billion people watched Live Aid.

Months earlier, Michael Jackson and Lionel Richie had written "We Are the World." All of it was a response to a famine in Ethiopia.

Almost nobody remembers who actually caused the famine. 🧵

[Jon Oringer @jonoringer]

sooo.. To match the inference speed and intelligence of a production-hosted Claude 3 Opus (or comparable 2026 frontier model), GLM-5.2 requires 8 NVIDIA Blackwell B200 or B300 GPUs running in FP8 quantization...

[Bob Loukas 🗽 @BobLoukas]

If the Bitcoin thesis is purely a SOV play, then in a world of AI capturing all the attention/capital, coupled with quantum fear, an outcome similar to this for the next 4 yr cycle would not be a big surprise, to me.

Would also align with a secular crypto bull market having peaked (as seen in alt's and bitcoin poor performance).

[Marc Andreessen 🇺🇸 @pmarca]

Interesting. (Founder of https://t.co/PY9jRPdR6L, creator of GLM AI models.)