标签：#t

The Quest Begins (The “Why”) I still remember the night our API started to sputter under a sudden traffic spike. Users were seeing 502 errors, the monitoring dashboard looked like a neon rainstorm, and I felt like I was stuck in a lobby waiting for the elevator that never arrives. We had a simple round‑robin load balancer sitting in front of three identical services. It worked fine when traffic was smooth, but as soon as a burst hit, one node would get overloaded while the others twiddled their thumbs. Honestly, I thought we just needed more servers. Throwing hardware at the problem felt like using a sledgehammer to crack a nut—expensive and messy. After a few frantic Slack threads and a lot of coffee, I realized the real issue wasn’t capacity; it was how we distributed the work. The balancer was oblivious to the actual load on each backend, treating every request like it was the same weight. That moment became my quest: design a load balancer that reacts to real‑time load, stays simple enough to operate, and doesn’t cause a reshuffling nightmare when we scale the cluster. The Revelation (The Insight) The breakthrough came when I read about least‑connections load balancing combined with a slow‑start period for new hosts. The core insight is deceptively simple: Send each new request to the backend that currently has the fewest active connections. Why does that work? Immediate fairness – If one node is handling long‑running requests, it will naturally have a higher connection count and receive fewer new ones until it catches up. Burst absorption – During a traffic spike, requests spill over to the less‑busy nodes instead of piling onto a single overloaded instance. Predictable scaling – When we add a new server, it starts with zero connections, so it gets a fair share of traffic right away—but we temper that with a slow‑start window to avoid overwhelming a cold host. Compare that to round‑robin, which blindly cycles through the list regardless of each node’s state. In

There's a school of voice-agent development that treats turn detection as something you buy, not something you build. Pick a streaming STT provider, let its end-of-turn logic decide when the user is done, and move on. For a lot of teams that's the right move — and if you're weighing the options, our breakdown of turn detection vs forced endpoints is the place to start. But some teams have already solved turn detection. They've tuned their own voice-activity detection over thousands of calls, they know their audio, and they trust their endpointing more than any default. For those teams, a streaming model's built-in turn logic isn't a feature — it's something to work around. What they want is narrower and faster: hand over a finished chunk of speech, get accurate text back, get out of the way. That's the case for bringing your own turn detection and pairing it with fast ASR over HTTP. Turn detection is an architectural decision, not a default Here's the framing that matters. In a streaming setup, the STT model is a participant in the conversation — it's watching the audio and deciding, continuously, whether the user has finished. That's genuinely useful when you want the provider to own that judgment. But it means the model is inserting its own decision between "user stopped talking" and "you get the transcript." If you already know the turn is over — because your VAD just fired — you don't want the model deliberating. You want it transcribing. Every millisecond the STT layer spends re-deciding a question you've already answered is latency you're adding for no benefit. So the decision isn't "which provider has the best turn detection." For these teams it's "who owns the turn boundary?" If the answer is you, then the ideal STT layer is one that does exactly one thing: turn a finished clip into accurate text, fast. Built-in vs. bring-your-own Built-in (streaming). The model reads tonality, pacing, and rhythm to detect end-of-turn — with Universal-3.5 Pro Realtime, aroun

You've got a recording and you want text back. For years that meant one thing at AssemblyAI: submit the file, wait for the job to finish, get a transcript. Async. It's reliable, it's cheap, and for a huge range of workloads it's exactly right. But "wait for the job to finish" is doing a lot of work in that sentence. If your file is two minutes long and your user is staring at a spinner, waiting is the whole problem. That's the gap the Sync API fills — and it's why "which transcription path" is no longer a two-way question. This post is about the two ways to transcribe a recording : async and sync. (If you're deciding between recorded and live audio in the first place — streaming versus the rest — start with our guide to real-time vs batch transcription , then come back here to choose between the two non-streaming paths.) The one-sentence difference Async transcription hands you a job: you submit audio, the work happens in the background, and you collect the result later by polling or via a webhook. Sync transcription hands you an answer: you POST a short clip and the transcript comes back in the same HTTP response — no job to track, no callback to wait for. Everything else follows from that. Async is built for throughput and depth on files of any length. Sync is built for speed on short files, when a person or an agent is waiting on the other end. How fast can each actually go? This is the question that usually settles it, so let's be concrete. Async processes the whole file and returns a single complete transcript, typically in seconds to a few minutes depending on file length and load. Crucially, it bills on audio duration ($0.21/hr on Universal-3.5 Pro), so a 30-minute file costs the same whether it comes back in 20 seconds or two minutes. You're optimizing for cost and completeness, not for the clock. Sync is built to return a transcript for a short clip almost immediately — roughly 134ms p50 — in one request/response, with no polling and no webhooks. It's price

Private inference for coding agents Discussion | Link

Array An Array is a collection of multiple values stored in a single variable. let fruits = [ " Apple " , " Mango " , " Orange " ]; Here, fruits contains three values. Why Do We Need Arrays? Without an array, you would write: let fruit1 = " Apple " ; let fruit2 = " Mango " ; let fruit3 = " Orange " ; Using an array: let fruits = [ " Apple " , " Mango " , " Orange " ]; This makes the code shorter and easier to manage. Array Index Each value in an array has an index. The index always starts from 0. Index: 0 1 2 ------------------------- Array: Apple Mango Orange Accessing Array Elements Use the index number to access a value. let fruits = [ " Apple " , " Mango " , " Orange " ]; console . log ( fruits [ 0 ]); console . log ( fruits [ 1 ]); // Output: Apple Mango Changing an Array Element You can update any value using its index. let fruits = [ " Apple " , " Mango " , " Orange " ]; fruits [ 1 ] = " Banana " ; console . log ( fruits ); // Output: [ " Apple " , " Banana " , " Orange " ] Finding the Length of an Array Use the "length" property. let fruits = [ " Apple " , " Mango " , " Orange " ]; console . log ( fruits . length ); // Output: 3 Adding Elements push() – Add at the End let fruits = [ " Apple " , " Mango " ]; fruits . push ( " Orange " ); console . log ( fruits ); // Output: ["Apple", "Mango", "Orange"] unshift() – Add at the Beginning let fruits = [ " Mango " , " Orange " ]; fruits . unshift ( " Apple " ); console . log ( fruits ); // Output: ["Apple", "Mango", "Orange"] Removing Elements pop() – Remove from the End let fruits = [ " Apple " , " Mango " , " Orange " ]; fruits . pop (); console . log ( fruits ); // Output: ["Apple", "Mango"] shift() – Remove from the Beginning let fruits = [ " Apple " , " Mango " , " Orange " ]; fruits . shift (); console . log ( fruits ); // Output: ["Mango", "Orange"] Looping Through an Array Use a "for loop" to print all elements. let fruits = [ " Apple " , " Mango " , " Orange " ]; for ( let i = 0 ; i < fruits . length ; i

You're building teams/workspaces on Supabase. A row belongs to a workspace, and a user can touch it if they're a member. Simple enough — until your policies start throwing infinite recursion and you have no idea why. Here's the trap and the pattern that avoids it. The setup A membership table, with RLS on it too (of course): create table workspace_members ( workspace_id uuid references workspaces , user_id uuid references auth . users , primary key ( workspace_id , user_id ) ); alter table workspace_members enable row level security ; create policy "read own memberships" on workspace_members for select using ( user_id = ( select auth . uid ())); The trap Now you scope a tenant table by asking "which workspaces is this user in?" — by querying workspace_members inside the policy : -- ⚠️ this can recurse create policy "members read projects" on projects for select using ( workspace_id in ( select workspace_id from workspace_members where user_id = ( select auth . uid ()) ) ); The policy on projects queries workspace_members , which has its own RLS policy, which re-evaluates… you've built a loop. The fix: a security-definer function Resolve membership with a function that runs with the definer's rights, so reading workspace_members doesn't re-trigger RLS. The policy becomes a plain IN check — no recursion: create or replace function public . user_workspace_ids () returns setof uuid language sql security definer set search_path = '' -- pin it: this is the safety bit as $$ select workspace_id from public . workspace_members where user_id = auth . uid (); $$ ; create policy "members read projects" on projects for select using ( workspace_id in ( select public . user_workspace_ids ()) ); Why it's safe: the function is read-only, returns only the caller's own workspace ids, and search_path = '' prevents search-path hijacking (the classic SECURITY DEFINER footgun). It never exposes another user's memberships. Do this for every tenant table (reads and writes — remember WITH CH

On June 9, 2026, Anthropic shipped Claude Fable 5, a model in a new tier that sits above Opus. I have been building on the Claude API for over a year, and this is the first release that made me stop and re-read my whole prompt stack before touching the model string. Here is what actually changed and what it means if you ship software. The short version Fable 5 is the public release of the Mythos line, the family that earlier in the year unsettled the security world with how well it found and exploited vulnerabilities. The version you and I get is the same underlying model with safeguards bolted on. Anthropic calls the safe one Fable and the unrestricted one Mythos, and only a small group of cyberdefenders gets Mythos. The numbers, for context: 1M token context window, 128K max output, knowledge cutoff January 2026. Priced at $10 per million input tokens and $50 per million output. That is double Opus 4.8 ($5 / $25). State of the art on nearly every benchmark they tested: 95% SWE-bench Verified, 80% SWE-bench Pro. Adaptive thinking is always on. There is no "disabled" mode. That last point matters more than the benchmarks. You do not tune a thinking budget anymore. The model decides. The pricing reframes the decision At $10/$50, Fable 5 is not your default model. It is your "this task is hard and getting it wrong is expensive" model. Opus 4.8 at $5/$25 remains the workhorse for most application traffic, and Haiku 4.5 at $1/$5 still wins on classification and routing. The way I think about it now is a three-tier ladder: Haiku 4.5 → routing, classification, cheap extraction Opus 4.8 → default for app traffic, agentic loops, coding Fable 5 → long-horizon agentic work where correctness pays for itself The "longer and more complex the task, the larger Fable's lead" framing from the announcement is the actual buying signal. A one-shot summarization does not justify 2x the cost. A multi-hour autonomous refactor that would otherwise need human correction might. The API surfa

I've been building AI agents for a while now, and streaming responses to a UI has always been the painful part. In previous projects I tried API Gateway streaming, Lambda response streaming, and even AppSync Events via an agent tool call to notify the UI. I also looked at adding my own WebSocket API through API Gateway, which requires managing $connect , $disconnect , and $default routes, storing connection IDs in DynamoDB, and posting messages back through @connections . All of these approaches felt like too much ceremony for what should be simple. That's when I found that AgentCore has built-in WebSocket support. The browser can just connect directly to the agent. No middleman. I built WearCast to demonstrate this functionality. This is an AI agent that helps you pick out what to where based on the weather. I've found this very helpfu when packing for a trip. The code for this application is public and you can find the full implementation on this GitHub repo . The problem with the traditional approach We usually build app applications as we do REST APIs User → REST API → Lambda → AI Service → Lambda → REST API → User Every message makes a round trip through multiple intermediaries. The response waits until the entire generation is complete, and then it all comes back at once. For a chat interface, this feels sluggish. Users stare at a spinner while the model generates hundreds of tokens they could already be reading. Even if you add Server-Sent Events or long-polling, you're still stitching together a real-time experience on top of infrastructure that seemed like overkill. I wanted something better. The architecture Here's what I ended up with: ┌─────────────┐ ┌──────────────────┐ │ React UI │─── JWT ─▶│ API Gateway + │──▶ Lambda (presigned URL) │ (Cognito) │ │ Cognito Auth │ └──────┬──────┘ └──────────────────┘ │ │ WebSocket (SigV4 presigned URL) │ ← No middleman! Direct connection → ▼ ┌──────────────────────────────────┐ ┌────────────────────┐ │ AgentCore Runtim

Find, research, and reach the right people Discussion | Link

Stripe and private equity firm Advent International have reportedly submitted a joint bid to acquire PayPal in a deal valued at approximately $53.4 billion. Reuters reports that the offer was submitted earlier this month and is backed by roughly $50 billion in committed bank financing. Under the proposal, Stripe and Advent would jointly own PayPal, […]

The US is getting closer to observing daylight saving time year-round. On Tuesday, the House advanced the Sunshine Protection Act on a 308 to 117 vote, which would turn clocks one hour ahead permanently, as reported by CBS News. President Donald Trump wrote on Truth Social in May that the bill would save the "hundreds […]

Sure, Bluetooth speakers are awesome to have for picturesque beach trips, camping weekends, and other outdoor adventures. But they’re just as good to have for some sprinkler time in the yard with the kiddo, or for blasting some tunes or a podcast while you make food. No matter the use cases you have in mind, […]

The FCC might be overstepping its bounds to open the door to broadcast TV monopolies.

Can a handful of engineers really do the work of an army of consultants? That’s the bet behind Ode with Anthropic — the joint venture dedicated to embedding forward-deployed engineers in enterprise firms, backed by Anthropic, Blackstone, Hellman & Friedman, Goldman Sachs and others. On this episode of TechCrunch’s Equity podcast, Rebecca Bellan sits down with Ode’s leaders Chris Taylor and Eddie Siegel, who founded Fractional AI, […]

Both 8BitDo and GameSir announced tiny gamepads designed to turn smartphones into Game Boy-inspired handhelds at CES 2026. GameSir released its Pocket Taco months ago, but the longer wait for 8BitDo's FlipPad has been well worth it. While the Pocket Taco may be a better fit for some players and devices, the FlipPad is a […]

Give every AI agent trusted context Discussion | Link

The feature was previously only available to paid subscribers.

New Delhi announced a $6.5 billion smartphone manufacturing program and a $13.3 billion semiconductor push to deepen India's electronics supply chain.

Similar to last time, What’s !important #15 is pretty stacked — read all about boundary-aware CSS, making grid lanes accessible, creating time-based web designs, fixing full-bleed CSS, improving customizable select, new web platform features, and more. What’s !important #15: Boundary-aware CSS, Time-based CSS, Full-bleed CSS, and More originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.