Vishal Talasani

let's go exploring

[Vishal Talasani @shellvish]

@RoboMcGobo @AgentKwosh Just an incredible chart all around

[Vishal Talasani @shellvish]

(1) hilarious graph

(2) it's incredible how consistent growth has been over the past 150 years. Despite massive innovations like trains, lightbulbs, cars, computers, internet, etc GDP growth is remarkably steady at ~2%/yr

[Vishal Talasani @shellvish]

Claude is currently offline

The last major outage was 2 working days ago, when all Anthropic services were down for ~10% of the workday

Anthropic, please raise prices!

[Vishal Talasani @shellvish]

Personally, I think Anthropic needs to raise their prices.

It's obviously bad that Anthropic silently throttled Claude Code, but you have to sympathize with their position.

Demand for their product is much higher than they can support.

They have 3 options:
1. Make no changes. Claude will continue to have serious reliability issues.
2. Degrade performance. Users will experience a worse product.
3. Raise prices, many users will churn to other providers.

Importantly, they must choose at least one of these paths, independent of how much more money they're able to raise. Anthropic is compute limited, and there are no obvious ways for them to increase the compute available to them in the short-term.

Option 1 is a non-starter, many of Anthropic's serious customers (e.g. enterprises) will churn as product outages continue to increase. Degrading performance is also bad, Anthropic is losing their lead as the cutting-edge provider, and users are very sensitive to any product regression.

Raising prices will price out many users from Claude Code, but will be net-better for enterprises, who prioritize reliability and performance. It will also help retain Claude's (small) edge in being the best coding model.

In practice, Anthropic is doing a mix of all 3 options, which imo is the worst outcome. They are silently degrading their product, restricting Claude Code to more expensive plans, and rolling back changes when users complain.

As a result, they now are faced with performance concerns ("is Claude Code worse today"), are rejecting lower-end customers, and have worst-in-class reliability.

It's painful to rip the bandaid off and just raise prices. But they need to do it at some point, and it's better to do it before their brand suffers.

[Vishal Talasani @shellvish]

Personally, I think Anthropic needs to raise their prices.

It's obviously bad that Anthropic silently throttled Claude Code, but you have to sympathize with their position.

Demand for their product is much higher than they can support.

They have 3 options:
1. Make no changes. Claude will continue to have serious reliability issues.
2. Degrade performance. Users will experience a worse product.
3. Raise prices, many users will churn to other providers.

Importantly, they must choose at least one of these paths, independent of how much more money they're able to raise. Anthropic is compute limited, and there are no obvious ways for them to increase the compute available to them in the short-term.

Option 1 is a non-starter, many of Anthropic's serious customers (e.g. enterprises) will churn as product outages continue to increase. Degrading performance is also bad, Anthropic is losing their lead as the cutting-edge provider, and users are very sensitive to any product regression.

Raising prices will price out many users from Claude Code, but will be net-better for enterprises, who prioritize reliability and performance. It will also help retain Claude's (small) edge in being the best coding model.

In practice, Anthropic is doing a mix of all 3 options, which imo is the worst outcome. They are silently degrading their product, restricting Claude Code to more expensive plans, and rolling back changes when users complain.

As a result, they now are faced with performance concerns ("is Claude Code worse today"), are rejecting lower-end customers, and have worst-in-class reliability.

It's painful to rip the bandaid off and just raise prices. But they need to do it at some point, and it's better to do it before their brand suffers.

[ClaudeDevs @ClaudeDevs]

Over the past month, some of you reported Claude Code's quality had slipped. We investigated, and published a post-mortem on the three issues we found.

All are fixed in v2.1.116+ and we’ve reset usage limits for all subscribers.

[Vishal Talasani @shellvish]

tokens are much less expensive than your time

[Vishal Talasani @shellvish]

My feed seems to be filled with Vercel haters. Where are my fellow Vercelcels at

[Vishal Talasani @shellvish]

Switching from local to cloud coding agents has been a major unlock for me.

My cloud setup today:

1. Instant bug resolution. Any bug immediately triggers an Agent, who will go through logs, code, and context to try to resolve the issue. All I look at afterwards is a PR with a clear description of what broke, a working recreation of the incident, and a reviewed fix.

2. High quality PRs. I review + merge ~90% of PRs within 5 minutes. All PRs get several reviews automatically, from security, code quality, test coverage, and UX perspectives. The agent automatically addresses these reviews. I only engage after all reviews pass. When reveiwing PRs, I only look at (a) screenshots of the new FE, (b) description of any new architecture, and (c) examples of BE changes.

3. Assign tickets from anywhere. Any new GH issue becomes a PR. Kick off agents from Slack. A trusted user files a bug or requests a feature? It's implemented immediately.

4. Proper FE testing. Cloud agents always take screenshots of their FE work, and analyze them to make sure they have the desired affect. Reviewer Agents also analyze those screenshots. This can be done locally, but it's much easier to reproduce consistently in Cloud.

[Vishal Talasani @shellvish]

@mohbii agreed!

a common issue I run into is CC will just break on web sessions, and will require multiple bumps over 5-60 minutes to restart it. this isn't captured in the status page, but is a real outage imo

[Vishal Talasani @shellvish]

I'm a big fan of Claude, and have become a heavy user of their Cloud product.

Lately though, its reliability has been a major issue.

The status page doesn't line up at all with my experience, I get many intermittent outages everyday.

Are others experiencing the same thing?

[Vishal Talasani @shellvish]

thinking about all the datacenters we could build here

[Vishal Talasani @shellvish]

As a fellow UChicago math major (‘18), I think this is mostly right, but there are many ways to end up with a math major!

A good friend of mine actually started in the Math 130s sequence (meant for students with no calculus background), decided he loved it, and ended as a math major. He now is full-time doing research as a math postdoc

My perspective is most high schoolers don’t have a good understanding of what they want to study. This is okay! College courses differ drastically from what even advanced high schoolers study (HS calculus is radically different from typical undergrad math)

It’s a credit to UChicago and other colleges that they give students the flexibility to try subjects, see what they enjoy, and then pursue that. Regarding the original physics major, my guess is they got similar value from a physics degree vs math (eg pursuing an industry where it just matters they’re experienced in thinking analytically), and physics resonated more with them

I disagree that students who are “only” 99th percentile can’t catch up to others in the major. But, they are confronted with what that major actually entails and if it makes sense given their goals

[Vishal Talasani @shellvish]

@thesamparr Great book, it makes you see NYC in a new light.

If you’re interested in Caro, I think you’ll love LBJ book 1 though.

The book starts with a history of rural Texas. It reads almost like a Western.

Whatever you choose, highly recommend jumping in. Caro’s process is incredible.

[Vishal Talasani @shellvish]

@0xTenkaIchi @AxiomExchange Yeah I think that's definitely a big part of it!

But I don't see it often discussed in tech circles, when people talk about other great YC companies

[Vishal Talasani @shellvish]

Is @AxiomExchange the best YC company from the last year?

Launched 6 months ago, and has earned $160m in revenue so far.

Currently earning roughly as much revenue as Brex, to put it into perspective.

Insane numbers, and not often talked about. Why is that?

[Riley @interchainriley]

appreciation post:

sam is stride’s first employee and a real-life 10x engineer

if using stride ever feels frictionless, it’s because sam spent the time sweating every edge case

an unsung hero of cosmos

happy 3 year anniversary at stride @sampocs13 🎉

[Vishal Talasani @shellvish]

At Cosmos EVM Day. The energy is here!

Haven't been to a Cosmos conference with this much excitement since Medellin in 2022

Hats off to @BPIV400 and @0xMagmar for making this happen 🔥

And great work @seb3point0 and the whole team for putting this together. Excellently done!

[Vishal Talasani @shellvish]

@0xTenkaIchi soon 👀

[Vishal Talasani @shellvish]

Apps should respect users, and avoid wasting their time.

It took us many hours, IBC upgrades, audits, edge case testing, and so much more to make Stride liquid staking 1 click.

Now we’re applying that same philosophy to Stride DEX.

[Vishal Talasani @shellvish]

@krishnanrohit I’m in the same boat!

Have switched to decaf coffee. imo it’s the same flavor profile, but fewer negative effects of drinking many cups

Are you trying to decrease your caffeine intake, acidity, or something else?

[Vishal Talasani @shellvish]

@ImmutableLawyer @KodiakFi @eatsleepyeet @BakerDAO420 Thanks a lot Joseph!

[Vishal Talasani @shellvish]

Stride's focusing on the basics. We're going all-in on Stride DEX.

@kodiakfi and @eatsleepyeet are excellent, and will continue to grow stBGT within @BakerDAO420

20% of profits will go to STRD buybacks. BakerDAO gets a leading LST that grows their product suite

Everyone wins

[Stride @stride_zone]

A governance proposal was just posted to authorize @BakerDAO420 to acquire Stride's stBGT.

This proposal will let Stride fully focus on Cosmos and Stride DEX.

Incubated by @kodiakfi and @eatsleepyeet, BakerDAO will continue to grow stBGT and will profit-share with Stride 👇🧵