SSR

Writing about AI coding tools and agents (Cursor, Claude, Grok, Codex), dev infra, and what works on real teams. No hype.

[SSR @ssraavi]

@emollick And the gap closed so fast that "which model" is now the least interesting question on most teams. The agent loop, the tools, and the review step around it decide whether anything ships.

[SSR @ssraavi]

https://t.co/e2XiZt02In

[SSR @ssraavi]

@elonmusk @SpaceX 700 flights is the real moat. Not the rocket, the reps.

Every landing is a data point most launch programs get a few times a year. Iteration speed beats brilliance, in rockets and in software.

Hoping India's private launch startups steal this playbook.

[SSR @ssraavi]

Typing speed stopped being the bottleneck for a lot of us the day coding agents showed up. The new bottleneck became knowing exactly what we want. Faster output straight from the brain sounds great, but most of my thoughts need a draft and an edit before they deserve to leave my head. The feature I'd want first is the undo button.

[SSR @ssraavi]

@paulg A round is a loan of time, not a trophy.

The teams that celebrate it hardest usually spend that time hiring to look like the company the deck promised, and suddenly burn is the roadmap.

The best use of a Series A is buying enough runway to prove you never needed the B.

[Dr prashant bhamare @dr_prashantsb]

A very significant video. Share with all the ladies you know ...

[Global Christian Relief @GC_Relief]

Stop going to church - or lose everything.

That's what the Kawde family in Chhattisgarh, India was told after their public baptism in February 2025.

They refused.

[Slack @SlackHQ]

When we launched Slack Code, partners and customers alike all had the same question: “Can I bring my own agent to the party?”

Good news, friends: the answer is yes. 🎉

Today, the Slack Code API is generally available, allowing your agents to work in code channels with diffs, previews, and plans. At any point along the way, your team can step in, add context, and approve changes — all with the permissions and admin controls you already know.

Start building: https://t.co/WyCxmeGwzT

[Gary Adams @GaryAdams713]

🤷‍♂️💯

[Logan Kilpatrick @OfficialLoganK]

Introducing Gemini 4 Carbon, it's a big jump over Argon which we released just 2 weeks ago.

Gemini 4 Carbon is SOTA on cyber, long horizon tasks, ranks #1 on Artificial Analysis V5, significantly improves on usability in domains like coding, and is more aligned.

[Claude @claudeai]

New in Claude for Legal: a gallery of ready-made artifacts for contract review, regulatory horizon scans, cite checks, and more.

Start from the built-in example, add your own documents and connectors, and share it with anyone across the business.

[SSR @ssraavi]

@paulg That's not a side process, it's an interrupt. Zero cycles spent polling for "do your homework", but his name on the bus fires an IRQ instantly. The handler for "clean your room" was just never registered.

[SSR @ssraavi]

@TVanWelsenes Depends on Device, Mode & Application usage. If its mobile, I would prefer middle. In desktop left. In tablet right.

[SSR @ssraavi]

Devs don't. Whoever signs the invoice does.

A single dev optimises for "did it ship before standup". Token spend only becomes a topic when 40 people run agents all day and finance asks why the AI line is creeping up on the cloud bill.

Bad YouTube topic, very real Slack thread. And outside the US, a $200/month plan is a different conversation entirely.

[SSR @ssraavi]

I'd push back a little. Coding isn't saturated, the easy 80% is.

The other 20% is the 12 year old monolith, the flaky infra, and the spec nobody ever wrote down. That's where sub-frontier models confidently ship the wrong fix and frontier ones still earn their price.

Discovery is a great bet. Just don't price enterprise coding like it's already a commodity.

[SSR @ssraavi]

Amazon is reportedly putting most Nova models on keep-the-lights-on mode, closing its AGI Lab and winding down the Q Developer IDE plugins.

Honestly, the sensible call. AWS never needed to win the model race. It needs every model that does win to run on Bedrock and bill through AWS.

The lesson for everyone else: a company with that much money and talent still couldn't make a me-too model people actively picked. If your differentiator is "we also have one", you don't have a differentiator.

Pick the layer you can actually win.

[☭ YASHWANTH ⚫💊 @Comrade_Yash]

Idhi yaaparam

[Sai @saidotdev]

even the guy who invented it doesn't qualify 😭

[SSR @ssraavi]

@owenthcarey Shalom

[SSR @ssraavi]

@elonmusk @poteto @bot 4.6 was with token efficiency, not available much in 4.7. Performance too