Sharath Challa

@_ARahim_ @bcherny only boomers fix typos in prompts. llms perfectly understand you even if you mistype.

@Brainformative Yep, the demo usually works. What breaks is handoff ownership. Once onboarding or support leaves the happy path, you need a clear owner, idempotent retries, and a shared state model; every edge case becomes manual triage. Where does that break first for you?

3/ They know the automation is just 10% of the work. The other 90%: • Client onboarding • Customization • Troubleshooting • Support (24/7) • Updates when APIs change • Managing expectations You can't snapshot that.

@0xCaitlin Congrats !! That manual layer is the symptom. The hard part is keeping the invoice state, ERP sync, and exceptions in one place. Most AR automation works on the happy path, then breaks on partial payments, credit memos, and human follow-ups.

WE RAISED OUR SERIES A!!! 🎉 🚀 Today, we’re thrilled to announce our Series A, bringing @FazeshiftHQs total funding to $22M. Accounts receivable is still one of the most manual, painful workflows inside modern companies. It lives in spreadsheets, email threads, and disconnected systems – and it slows down how businesses get paid. That’s exactly why we built Fazeshift: an AI-native platform that automates and scales AR, helping finance teams move faster, operate with clarity, and unlock cash flow. Since we kicked things off in 2024, the team has executed relentlessly: ✔️ 12x revenue growth ✔️ 8 unicorns now using Fazeshift ✔️ Team grew from 2 → 26 ✔️ Opened our SF HQ and NYC office This milestone is a reflection of the people behind it, and it gives us the ability to invest more aggressively in our AI agent’s capabilities. To our investors, thank you for your belief and partnership. This financing ranks well above the typical Series A, reflecting strong investor conviction in the shift toward AI-native finance infrastructure. F-Prime, @GradientVC , @ycombinator, @0xpioneer_fund, Phoenix Fund, Wayfinder, @RitualVC , @2048vc , The Graduate Syndicate, Vulcan Partners, and incredible angels including Sandor Palfy, Kulveer Taggar, Richard Aberman, Terrence Rohan, Rob Biederman, Mike MacCombie. To our customers, thank you for trusting us with something so critical. You’re trusting us with how your business gets paid, and we don’t take that lightly. And to @0xTimmyG and our team, it’s an absolute privilege to build alongside you. We’re just getting started. This raise isn’t the finish line, it’s fuel. Let’s go do this thing! Read the full story from @bayareawriter in @crunchbase (link in comments).

@cursor_ai The tricky part is not fixing the failing test; it is classifying whether the failure is code, env, or test flake. If the agent skips the repo-scoped context and a confidence gate before PRs, it can loop on the same flaky path forever. How are you separating those?

Cursor can now automatically fix CI failures. Set up always-on agents that monitor GitHub, investigate root causes, and open PRs with fixes.

@tibo_maker That “content → products” handoff is where most systems slip. SKU/variant canonicalization + stale feed data will quietly break mapping even when the content is good. The real work is the sync layer and reconciliation rules. How are you deduping variants?

I am very proud of this one. ecommerce stores often have 10k+ products. the hard part is not just creating SEO content, it's building the entire conversion funnel, from content to products so we shipped product catalog integration in Outrank content can pull your catalog, mention relevant products naturally, generate in-context visuals, and push product pages without turning the article into a sales pitch this is Outrank becoming the best SEO tool for e-com

@ayushjaiswal Coinbase’s one-person teams idea makes sense, the code part gets cheaper. The bottleneck shifts to context handoff, review loops, and evals. When design/product/eng collapse, quality drifts unless someone owns spec drift. How do they keep that tight?

Coinbase is testing one person teams combining design / product / engineering. This is a huge win for Claude & Codex. Work is changing so quickly.

Love this tiny UX detail from @supabase, adding a "Last used" badge on OAuth login options. - Prevents duplicate signups. - No more guessing which email or provider lastly used. - Just makes sense. Every app should copy this. Small touch, big impact. #UX #OAuth #DevTools

@aribk24 Vocaldesk (AI Phone agent) is up for sale! vocaldesk.co For some background on why I’m selling, check out this post: reddit.com/r/SaaS/s/vHl7L…

I have $500k cash to spend on random apps + sites If you built something cool and want to sell it drop a link below

AI

shadcn/ui is like 'Everything Everywhere All At Once' for designers & frontend devs. @shadcn

Are paid ads useful? I want to try but seeing lot of people complaining about the conversion rates.

Other than text and media generation, where else AI is used in adtech?

Beyond excited to have stumbled upon @gummy_search

Question ‘why’ is more harder to figure out than ‘what ‘ and ‘how’.

If you're developing on macOS with Docker Desktop, give @OrbStack a try. The startup times and resource usage are impressive, and best of all, I freed up a significant amount of disk space after uninstalling Docker Desktop.

Short courses in @DeepLearningAI are real gold.

@dannypostma Is this some secret?

Tech as "just a means to an end" is dangerously short-sighted and misses the mark when it comes to AI. As AI evolves, it's not just a tool — we're shaping a new kind of intelligence. Dismissing it as just another tool could blind us to the risks of AI overtaking us. I strongly believe that this will happen in the next few decades. Look at the hockey stick growth of AI over the past few decades. As they say in business, going from 0 to 1M takes years, but growth from there on is automatic. This was and is true for tech.

A few weeks back I saw @X generating AI summaries on the profiles. Where did they go now?

@milan_milanovic AFAIK, engineering teams never talk about customers in scrum or it is the least discussed topic.

What is good with Scrum: - Working iteratively - Focus on customer What is not good: Everything else, including: - Scrum master role - Story points - Velocity - Feature factories - And more What is your take?