Will

There are verifiable rewards everywhere for those with eyes to see

@rohitdotmittal @jxnlco open source is the way they will win.

Codex has a short window to catch the Claude train. OpenAI has no money advantage anymore. 95% of the videos by influencers on my timeline are about building systems with Claude. Codex is no where to be seen. This is in spite of the fact that Codex with 5.4 is an equal or a better product.

@thdxr this combined with each environment getting access to a git repo with a commit made at the end of each prompt (for full rewind)

we've been experimenting with getting rid of the bash tool agents can write js fine which can do what bash can (though some gaps with things like git) and is more cross platform and then could run that in this

@ryancarson @linear isn’t this what symphony is?

The company that actually builds the agent-first code factory is going to be worth hundreds of billions of dollars. No one has cracked it yet. It can't be the model labs because then you're tied to one model. I'm hoping a company like @linear will do this. I'd happily pay thousands of dollars a month for that (+ the token cost). Basically, we need SDLC 2.0 for the agent age. (Also, the right solution can't rely on gh - we need whoever does this to completely replace it as well.)

@emollick it’s taking openai months to implement hooks into codex.. still haven’t done it

Still no Claude Cowork competitor from any other lab yet. On one hand, its been six weeks. On the other, its been six weeks for companies that say that all their code is being written for them by AI.

We granted all current Claude Pro and Max users $50 in free extra usage. This credit can be used on fast mode for Opus 4.6 in Claude Code. To use, claim the credit and toggle on extra usage on claude.ai/settings/usage. Then, run `claude update && claude` and `/fast`. Enjoy!

@yibie @theo only solves the CLI side - still blocked by browsers and multiple backend services in different states

@theo There is a project for this context: supacode.sh

@buccocapital @dalibali2 Browser use agents that pull data out without APIs will stop this from happening

@dalibali2 You have to believe they can somehow lock down the context/maintain data gravity to keep their pricing power. TBD

Anthropic uses Workday. OpenAI uses Slack. It’s incredibly clear to anyone with half a brain that nobody is vibe-coding critical infrastructure. It is genuinely the lowest EV activity you can do. That bear case is dead (to I think most sensible investors). BUT, there are others. Here’s my remaining set of bear cases for SaaS, stack ranked. If you’re going to invest in SaaS you should be aware of all these and have a very strong POV on how impactful they are and the timeline on which you think they will (or won’t) play out 1. Platform differentiation trends toward zero, hurting CAC as each customer/upsell becomes a knife fight with multiple competitors. We were already trending toward every platform offering every app, AI just made it easier 2. Value will accrue to the agentic layer sitting on top of the system of record (SOR). Even with all their context, a SOR still lives in a silo 3. Investor sentiment becomes a structural headwind as revenue quality/business model is de-rated 4 . AI-native startups will deliver tremendous value at better prices, eating the incremental LTV of incumbent customers 5. As Agents do more work, seat-based revenue will decline. 6. Legacy SaaS will struggle to transition from seats to outcomes 7. Diminished pricing power due to decreased differentiation and lock-in takes away yet another growth lever 8. Gross margins will deteriorate because AI revenue is structurally more expensive which hurts the value prop of the business model 9. Decreased organic traffic due to LLM adoption increases CAC 10. Competition for scarce AI talent increases SBC/opex faster than revenue as incumbents fight to keep A talent from AI-native companies

@beyang beans or tk?

Gotta clean house for what's coming next

@peekknuf @GergelyOrosz this is the way.

@GergelyOrosz tech interview as in "let's have a chat about tech" followed by "show me something you've built" finished by "walk me through it" sprinkled with a couple of whys here and there This instead of braindead algo questions? You tell me

Calling it now: Onsite days and onsite loops will soon be back. Exactly like they were up to 2020. For mid-sized and above companies that do hybrid work. First screening: phone/video. Then onsite day with 4-5 interviews, back-to back. Hard to see this ending any other way.

@DarrelFrater I’ve already pitched before but very keen to go again 😂 DM

Any founders interested in pitching my friends at Stout Street Capital? They are a Denver based VC firm that believes great startups can be built anywhere. 💰 Check Size: $250K Initial, up to $1M Follow On 📈 Stage: Pre-Seed & Seed Comment "DM". Happy to get you connected.

@DarrelFrater interested

Any founders interested in pitching my friends at Geek Ventures tomorrow? 💰 Check Size: $200K - $800K 📈 Stage: Pre-Seed & Seed 📍 Geography: Worldwide 💼 Industry: Agnostic, B2B SaaS They focus on "Immigrant Founders". Comment "Interested". Happy to get you connected.

@eastdakota when will r1 be available on workers AI?

Turning up mothballed nuclear power plants was never going to be the answer. There is *so much* opportunity to improve the efficiency of AI. #DeepSeek did merely two clever optimizations. There’s a lot more to come and I’m glad they’ve refocused AI research on efficiency.

@rickyrobinett @CloudflareDev @OpenAI Is there any ETA on supporting Deepseek R1 through Workers AI?

I'm a kinesthetic learner 🎓, so I often build basic versions of rad AI tools to grasp new concepts. This week's project is (unsurprisingly) giving an AI browser access. Sneak peek with @CloudflareDev Browser Rendering + @OpenAI GPT-4o 👇 Full video and code this weekend!

we are currently in the intelligence arbitrage era: ai competence is wildly uneven rn—most people have no clue how to use it effectively. exploit the gap while it exists, because markets will equilibrate faster than ever now.

@nico_mannucci @codyschneider there’s a few people doing this - trychirp.com is one

Instead of an ai agent doing the whole workflow, ai agents should just assist people in doing more personalized outreach.

everything ive heard is that the churn on ai sales agents is more painful than an emergency root canal and there's 17+ funded yc companies doing it such a pitiful use case think bigger

Jia Tan found a way to get paid to work on open source. Why can't you

@t3dotgg Loved losing all my data today without warning.

Holy shit Insomnia actually just wipes all the data you had locally with the latest update if you refuse to sign in with an account 😳 Didn't think they could one up Postman's egregious user hostility, I'm almost impressed

@ImSh4yy Just implemented the exact same thing, except with openresty. Tried cloudflare for saas unsuccessfully as well.

Story of my embarrassing launch. 🥹 Right after I made Draftdock public, I started getting many messages from users who needed to configure their domains to their new blogs. I was using Cloudflare for SaaS as it made verifying and generating SSL certs for vanity subdomains easy. So, I looked into their system and noticed that, yes, most domains are failing. I tried everything I could think of to see if I could resolve this, tested with different domains, changed my implementation, and reached out to their community and other founders, but nada, it didn't want to work. At this time, it's almost 6 - 7 hours into the launch, I'm getting a ton of signups, and nobody can get their domains to work, and it's getting quite embarrassing and frustrating. So, my developer nature kicked in, and I decided to entirely ditch Cloudflare and build my own setup instead. After some research, it turned out Caddy was the perfect tool for this job. It has a neat little feature called `on_demand_tls,` which generates SSL certs on the fly. It also lets you specify a callback URL to verify the domain name. Combining these two, a tiny endpoint I built and their S3 storage plugin, made a fantastic combo for verifying domains, generating SSL certs, and persisting the certs on S3. This was my first all-nighter in a long time, but I got a solid setup up by 7 in the morning, and it worked like a charm. After a quick powernap, I pushed the fix to production, and we finally had a working product 24 hours after launch. On the bright side, the verification process is much easier and faster now; things are more reliable and cost me nothing to run.

@dannolan Yeah infinitely better. Still early days though, kernel panics if you run it on Intel and put the machine to sleep with containers running.

docker on mac absolutely sucks, this completely replaces it and rules orbstack.dev