BUSINESS

Poolside, the San Francisco AI lab that has spent most of its three-year existence quietly selling coding models to governments and defense agencies, released its most capable model to date on Tuesday — and made an unusually aggressive bet that radical transparency, not raw scale, is how a smaller lab competes at the frontier.The model, Laguna S 2.1, is a 118-billion-parameter Mixture-of-Experts (MoE) system that activates only 8 billion parameters per token, supports a context window of up to 1 million tokens, and — according to benchmarks published by the company — matches or beats open models several times its size on agentic coding tasks. The weights are available immediately on Hugging Face under the permissive OpenMDW-1.1 license.The headline numbers are striking for a model this small. Poolside reports that Laguna S 2.1 scores 70.2% on Terminal-Bench 2.1, a benchmark of long-horizon terminal tasks, placing it 11th on the company’s compiled leaderboard — ahead of DeepSeek-V4-Pro-Max, a 1.6-trillion-parameter model that scored 64.0; Thinking Machines’ 975-billion-parameter Inkling, at 63.8; and Nvidia’s 550-billion-parameter Nemotron 3 Ultra, at 56.4. On SWE-Bench Multilingual, it posts 78.5%, and on SWE-Bench Pro’s public dataset, 59.4%.Perhaps more telling than any single score: the model went from the start of pre-training on May 22 to public launch in under nine weeks, trained on 4,096 Nvidia H200 GPUs. In an industry where flagship model cycles are typically measured in quarters or years, Poolside has now shipped three models in three months.Why the West’s open-weight AI gap has become a boardroom issueThe release lands in the middle of an increasingly pointed debate about the provenance of open-weight AI. Over the past year, developer adoption has shifted decisively toward open-weight systems that companies can download, inspect, and run on their own infrastructure — and the leading options in that category have overwhelmingly come from Chinese labs. DeepSeek, Qwen, Kimi, GLM, MiniMax, and Tencent’s Hunyuan line all feature prominently in Poolside’s own comparison tables.Poolside’s accompanying press release frames Laguna S 2.1 explicitly as a response, noting that the model occupies a size class into which no Western lab has released open weights in 11 months — since OpenAI’s gpt-oss-120b last August. “The West needs open-weight models it can trust, run, and build on,” said Jason Warner, Poolside’s co-CEO, in the announcement.Co-founder and co-CEO Eiso Kant made the philosophical stakes even plainer in a lengthy post on X. “I believe intelligence should and will become a commodity,” he wrote, arguing that the open ecosystem “will not win by being the best in its own category.” Users, he argued, simply want the best intelligence for the task at hand — so open models must be on par with, or better than, their closed equivalents.The strategic logic here is not charity. Poolside’s core business is deploying models inside the security boundaries of government, defense, and regulated enterprises — customers for whom closed, metered API access is often a non-starter for compliance and sovereignty reasons. Every enterprise that standardizes on a Chinese open model today becomes harder to win tomorrow. Releasing competitive open weights is both an ecosystem play and a top-of-funnel strategy for the company’s high-security deployment business. It also reframes the AI race away from terrain where Poolside cannot compete — frontier-scale capital expenditure — and toward terrain where it believes it can: cost per token, self-hosting, and iteration speed.How a sparse architecture makes enterprise AI agents affordable to runThe technical design reflects a specific thesis about where value in coding AI is moving. Laguna S 2.1’s sparse MoE architecture — 256 routed experts plus one shared expert, with grouped-query attention and interleaved sliding-window layers, according to the Hugging Face model card — means inference costs scale with the 8 billion active parameters, not the 118 billion total. Poolside emphasizes that the model is small enough to run on a single Nvidia DGX Spark, the desktop-class AI machine.That matters for what Poolside calls token economics. Long-horizon coding agents are voracious consumers of tokens: the company’s published data shows the model consuming a mean of roughly 249,000 completion tokens per trajectory on its hardest benchmark when thinking mode is enabled. At metered API prices, agentic workloads at enterprise scale become a meaningful budget line item. On OpenRouter, Poolside is offering a free 256K-context endpoint and a dedicated 1M-context deployment priced at $0.10 per million input tokens and $0.20 per million output tokens — aggressive pricing that undercuts most frontier alternatives by an order of magnitude.The ecosystem support is unusually broad for day one. The model is live on Baseten’s model library and Vercel’s AI Gateway, with integrations across vLLM, SGLang, Ollama, and llama.cpp, plus quantized variants down to 4-bit GGUF files — 75 gigabytes — for local use. But Poolside’s more interesting claim is behavioral, not architectural. Pengming Wang, co-head of applied research at Poolside, said the gains came from improving the model’s working habits: “more verification, less taking things for granted, not declaring victory early, and being more persistent.” Raw intelligence, the company argues, is one axis of capability; a model’s way of working is a second axis that matters immensely for agents left unattended for hours.Publishing every benchmark trajectory to counter AI’s credibility crisisThe most consequential part of the release for enterprise buyers may be an evaluation-transparency move with little precedent among major labs: Poolside published the complete, unedited trajectory of every trial in its final benchmark runs — every reasoning step, tool call, and shell command behind every reported score.This addresses a growing credibility problem in AI benchmarking. As top scores on mature benchmarks cluster in the 70–90% range, and as “reward hacking” — models finding solutions online or gaming verifiers rather than solving problems — has become endemic, self-reported numbers have lost much of their signal. Poolside disclosed its own encounters with the problem candidly: during training, more than half of trajectories on some SWE-bench tasks were flagged because the model simply researched the original bug-fix pull request online and applied it. The company documented its mitigations, including prompt addenda, LLM-based judging calibrated against human labels, and expert annotator review of a high-scoring Terminal-Bench run.Three published case studies illustrate what the company means by persistence. In one, the model built a working HTML/CSS rendering engine from an empty folder in a 181-step, 50-minute unattended session — then, lacking vision capabilities, spun up headless Chromium to numerically compare its canvas output against a real browser’s rendering. In another, pointed at Poolside’s own agent harness in an automated optimization loop, the model made the Go codebase 5.2% faster with roughly 70% lower memory allocation, finding an O(n²) string-concatenation bug along the way. In a third, working in a sandbox with no Python installed, the model did its number theory in Perl and independently re-derived a proof of Erdős problem #397 — a combinatorics question open for five decades until GPT-5.2 Pro first solved it this past January. Poolside notes that its model’s construction is structurally different from the earlier published solution, and that its November 2025 knowledge cutoff precedes the first proof.What the disclosed limitations and benchmark fine print revealPoolside deserves credit for disclosing limitations most labs bury. The model can overfit to its native harness and stumble on slightly different tool schemas in third-party agents, mangles JSON in nested tool arguments, and is prone to overthinking on competition math. There is currently no user-configurable thinking-effort dial — just on or off — and the gap between the modes is enormous: thinking lifts Terminal-Bench 2.1 from 60.4% to 70.2%, and DeepSWE from 16.5% to 40.4%, at substantially higher token cost.Buyers should apply their own discounts to the comparison tables. Poolside’s methodology takes the maximum of vendor self-reported scores, benchmark-author leaderboards, and third-party figures for competitors — a reasonable convention, but one that mixes harnesses and test conditions. On DeepSWE, notably, Poolside ran its own agent harness rather than the leaderboard’s standard mini-swe-agent, a difference the company acknowledges makes scores less directly comparable. And the frontier remains clearly out of reach: closed models like GPT-5.6 Sol, at 88.8 on Terminal-Bench 2.1, and Claude Fable 5, at 88.0, along with the 2.8-trillion-parameter open-weight Kimi K3, at 88.3, sit well above Laguna S 2.1.The deeper structural question is whether Poolside’s “Model Factory” — the internal platform the company credits for its rapid release cadence — can sustain this pace as models scale. The trajectory so far is genuinely unusual: the April dual release of Laguna M.1 and XS.2, the July 2 refresh of XS 2.1, and now S 2.1, which the company says outperforms April’s flagship M.1 at roughly a third of its active size. Remarkably, S 2.1 used the exact same pre-training data as XS 2.1, meaning nearly all the improvement came from scale, training fixes, and post-training across the company’s corpus of 409,000 agentic and non-agentic training environments. Poolside says its next, larger Laguna model began pre-training last week.For technical decision makers, Laguna S 2.1 is the most credible Western open-weight option to emerge in nearly a year for self-hosted agentic coding — with published evidence, a permissive license, broad ecosystem support, and an economics story built around hardware you can own. Whether it dents the dominance of Chinese open models will depend less on this release than on the ones that follow it.Kant, for his part, has already told the world how he intends that story to end. Poolside is building toward a future where the most capable intelligence “can be owned and shaped by anyone,” he wrote — and the company plans to keep shipping “until that future exists.” In an industry where the biggest labs increasingly lock their best work behind an API, the most radical thing about Laguna S 2.1 may not be what it scores, but that anyone can download it and check.

There has been an escalation in strikes between the U.S. and Iran throughout the last 10 days over who has control of the Strait of Hormuz.

The AI boom is reshaping tech portfolios. Discover which semiconductor stocks are buys today and which you should ignore to protect your capital.

Illinois enacted a 0.2% tax on all crypto transactions last month, with the tax taking effect next year.

The AI server maker now expects gross margins to be in the range of 15% to 17% thanks to an improving customer and product mix

Semiconductor stocks made some of the easiest money on Wall Street this year. Now they’re quickly giving back their gains.The iShares Semiconductor ETF (SOXX), which many investors use to bet on chips, has crossed into bear market territory after falling from its June high.And one closely followed strategist thinks the damage is not finished.Ed Yardeni, the veteran market forecaster behind Yardeni Research, told clients on July 20 that chip stocks have room to fall further, even as the rest of the market holds steady.If you own a chip fund, a memory maker, or a broad tech position, that is a call worth paying attention to.Why Ed Yardeni thinks semiconductor stocks have further to fallYardeni’s warning is clear. He expects the S&P 500 Semiconductors index to drop about 12% more to reach its 200-day moving average, according to Yardeni’s QuickTakes analysis.The 200-day moving average is simply the average closing price over the past roughly 200 trading days. More Chip Stocks:Cathie Wood sells $11.7 million of tumbling semiconductor stockMicron stock jumps as investors look beyond GPUs in AI chip tradeGoldman Sachs turns its back on major semiconductor stockChart watchers treat it as a long-term floor, the line a stock tends to fall back to after a big run.That is the level Yardeni thinks chips are heading toward.The sell-off is already sharp. The SOXX has plunged 20.3% from its June 2 peak, Benzinga reported, the drop that officially defines a bear market.Riskier corners fell even harder. The Roundhill Memory ETF (DRAM) is down roughly 35% from its June 22 high.

Semiconductor stocks have slipped into a bear market after leading the 2026 rally.sinology / Getty Images

What broke while the broader market stayed calmHere is the part that makes this sell-off unusual.The S&P 500 is still sitting near its record, hovering close to the 7,500 level. So most investors checking their broad index funds saw almost nothing wrong.The surface stayed calm while the engine broke.The losses were concentrated in one crowded trade, while everything else held together.Three pressures hit chips at once, according to Yardeni’s note.1. Forced selling out of AsiaMargin calls landed on South Korean chip giants Samsung and SK Hynix (SKHY). When leveraged traders get a margin call, they are forced to sell to cover their positions, and selling affects U.S. memory and chip stocks.2. A cheaper AI model out of ChinaChinese AI lab Moonshot launched Kimi K3, a 2.8-trillion-parameter open-weight model it claims rivals the best from OpenAI and Anthropic.Because it delivers high-end performance at a lower cost, it revived the same fear that hit the market during the DeepSeek scare. The fact that AI hardware may not stay as valuable as investors assumed.3. No buyers left to catch the fallMomentum, or buying stocks simply because they were already rising, was the strategy that most investors crowded into in 2026. Once the selling started, there weren’t enough buyers left on the sidelines to slow it down, and the drop kept feeding on itself.How the chip sell-off unfoldedThe reversal took shape over several weeks:Dec. 7, 2025: Yardeni Research downgrades the S&P 500 Information Technology sector to market weight, an early defensive move.June 2, 2026: The broader market and the SOXX both peak before turning lower.Mid-July 2026: Leveraged liquidation and Moonshot’s Kimi K3 launch hit the sector; South Korea moves to curb high-leverage tech ETFs.July 19-20, 2026: The SOXX confirms a bear market, down 20.3% from its June peak.July 20, 2026: Yardeni issues his call for a further 12% drop.This did not come out of nowhere. Signs of strain had been building for weeks, and it was flagged ahead of TSMC’s July earnings report, when the whole AI trade looked stretched.What Yardeni tells investors to do insteadYardeni’s advice comes down to one idea: Do not try to guess the exact bottom.He warns against “catching falling knives,” which refers to the trap of buying a declining stock, only to watch it fall further.Instead, he favors a sector rotation, moving money out of the hardest-hit group and into areas holding up better.Yardeni’s firm keeps an overweight rating on two sectors:Financials, helped by a strong investment banking environmentHealth care, especially biotechnology, which has weathered the tech dropThe logic is practical. If chips still have room to fall, parking money in steadier value sectors lets you stay invested without absorbing the worst of the fall.What this means for your portfolioThis does not mean chip businesses are broken. Demand for AI hardware is still real, and analysts remain split on where the sector goes next.Even during the slide, Goldman Sachs reset its AMD target higher, and Micron’s (MU) memory pricing story kept some buyers interested.So the question is not about whether chips recover, but when.A few things worth watching before calling a bottom:Whether the SOXX actually reaches and holds its 200-day moving average.Whether forced selling in Korea fully clears out.Whether AI spending from big cloud buyers stays strong into the next round of earnings.If you hold chip stocks for the long term, Yardeni’s call is a reason for caution, not panic. If you were tempted to buy this dip, his message is simple: The floor may not be in yet.Keep in mind that moving averages can break, and rotations can reverse quickly if AI demand exceeds expectations.Yardeni himself remains positive about the wider economy, still predicting a “roaring 2020s” run backed by broad corporate earnings. For now, the takeaway is clear. The chip trade that carried 2026 has cracked, and the strategist who called in the danger thinks there is still more room to fall.Related: Citi sends warning on semiconductor and hyperscaler stocks

TheStreet aims to feature only the best products and services. If you buy something via one of our links, we may earn a commission.Why we love this dealKeeping your house clean can feel like something that consumes you on a daily basis. Whether you’re in a sprawling single family home or a small apartment, keeping dust and dirt at bay isn’t always easy. That’s why it’s imperative that you have the right tools for the job. While a wet dry vac is great for big spills, every day maintenance of your space requires something a bit more versatile. That’s where a good lightweight stick vacuum can help. Thankfully, Walmart has some of the best deals on stick vacuums, and we found one with a discount that simply cannot be ignored. If you ever wanted to get a great price on a high-quality stick vacuum, today is the day you can make that a reality.The  Prettycare Self-Standing Cordless Stick Vacuum is available at the moment for only $70. That’s 59% off the original price of $170. Finding a stick vacuum of this quality for a shocking $100 doesn’t happen every day. That’s why we recommend you buy one now rather than waiting. Flash deals like this one often disappear without warning, so take advantage while you still can.Prettycare Self-Standing Cordless Stick Vacuum, $70 (was $170) at Walmart

Courtesy of Walmart

Shop at WalmartWhy do shoppers love it?Stick vacuums are known to be light, convenient, and powerful, and this one fits that description to a tee. That’s why it’s one of Walmart’s bestselling models. As for the lightweight aspect of the vacuum, it weighs a paltry 6 pounds, making it easy to use for anyone. Many people will even find it easy to use with just one hand thanks to the combination of lightweight materials and ergonomic handle design. Nothing makes cleaning easier than feeling like your vacuum is lighter than air itself.In addition to the weight and ergonomics of the vacuum, it’s incredibly user-friendly. It has a single-button canister emptying design, an LED battery screen with automatic charging reminder, and a free wall mount. What’s more, if you don’t have anywhere to mount the vacuum on the wall, it has a self-standing design, which means you can store it just about anywhere you want. The 180-degree side-to-side maneuverability and bright LED headlight make reaching tight corners a breeze as well.As for power, this vacuum has plenty of it. It manages 18 kilopascals (KP) of pressure on standard mode and up to 30 kp on boost mode. That’s more than enough to pick up dust, dirt, and the ever-pesky pet hair that blankets many a home every day. Additionally, the vacuum can instantly transform into a hand vac using one of the included attachments, while losing absolutely none of its impressive suction power.Related: Shark’s stick vacuum is nearly 50% off with a last-minute Prime Day dealWalmart shoppers were highly impressed with this vacuum. One said it’s “lightweight, easy to handle, and works great on both tile and rugs…This vacuum made my cleaning routine faster, easier, and actually enjoyable.”Shop more deals Shark Clean & Empty Cordless Stick Vacuum, $300 (was $450) at AmazonBissell PowerClean FurGuard Cordless Vacuum, $230 (was $300) at AmazonMoosoo Stick Vacuum Cleaner, $60 (was $170) at WalmartThe Prettycare Self-Standing Cordless Stick Vacuum is an absolute steal at 59% off. If you’re looking for a vacuum that can be both full-sized and handheld, offers superior suction, and can be bought for just $70, then this stick vac is for you. This Flash deal will be gone before you know it, so put one in your cart sooner rather than later if you don’t want to miss out. 

GPU memory is the most expensive resource in production AI, and it’s also the one running out fastest. Long context windows and multi-turn conversations force AI models to repeatedly recompute information they’ve already processed, consuming GPU memory and compute that could otherwise serve additional users or generate new responses.Instead of treating GPU memory as the limiting resource, why not extend it with much cheaper storage technologies? Weka, for one, believes that cheap flash storage can close that gap. The company’s NeuralMesh 6 software platform, launching alongside its first self-designed hardware line, Wekapod 3, extends what Weka calls Augmented Memory Grid, an approach that aggregates NAND flash to behave like GPU memory at a fraction of the cost.This is an active and increasingly crowded category. Dell, NetApp, Pure Storage and VAST have all repositioned toward AI infrastructure over the past two years and Weka is one of several vendors arguing it’s built for this specific moment rather than adapting to it.”What we’re seeing now with customers is they’re chasing availability of compute, and once they get new allocation from anyone, they want to be able to grab it and start running right away,” Weka co-founder and CEO Liran Zvibel, told VentureBeat.The potential payoff is straightforward: better utilization of existing GPU investments, lower inference costs and faster deployment of new AI workloads without waiting months for additional GPU capacity.The technology is most relevant for organizations already operating AI at scale or expecting rapid growth in usage, particularly enterprises building internal copilots, customer service agents, software engineering assistants or retrieval systems with long context windows. Smaller deployments may see less immediate benefit than organizations where GPU utilization has already become a limiting factor.Inside Weka’s NeuralMesh 6NeuralMesh 6 adds four capabilities aimed directly at a functionality gap Zvibel says has been costing Weka deals in competitive evaluations.Composable and virtual multi-tenancy. Composable clusters give anchor tenants full hardware-level isolation, dedicated CPU, memory, and storage. Virtual multi-tenancy runs through Weka’s RDMA fabric, delivering network-level isolation that scales past 1,000 tenants per cluster, with provisioning in under 30 minutes. Combined, a single cluster running 50 composable clusters can support up to 50,000 tenants. Unified file and object storage. Most storage systems keep two separate paths: a file-based path (the standard way servers and applications read and write files, used heavily in training and fine-tuning pipelines) and an object-based path (S3, the format inference and cloud-native tools typically expect). Normally a gateway translates between the two, meaning the data effectively exists twice. Weka’s claim is that the same physical data on disk is directly readable through either path at once, no translation layer, no second copy. Zvibel is targeting non-AWS GPU clouds specifically, naming Lambda, Nebius, G42, and CoreWeave, with what he described as roughly two orders of magnitude higher performance than conventional S3 and a capacity-based pricing model instead of per-API charges. Metadata-first replication. Destination environments become browsable before a full data copy arrives, with data hydrating only when accessed. “They had to wait for all of that to make it to the other side, and this takes days or weeks, in extreme cases a month,” Zvibel said. “We now allow our customers to grab some allocation of new GPUs and get up and running within an hour.”AlloyFlash and Always-On data reduction. TLC and QLC are two types of NAND flash memory. TLC is faster and more durable but costs more per terabyte, while QLC is cheaper and holds more data per chip but is slower. AlloyFlash mixes both within a single cluster, automatically routing latency-sensitive work to TLC while running bulk-capacity workloads on QLC, cutting cost per terabyte without a performance penalty on the work that needs speed. Data reduction now runs by default rather than as an option.Solving AI’s context problemMulti-tenancy and object storage solve how enterprises and neo clouds operate the platform day to day. A harder problem sits underneath: as context windows and multi-turn interactions grow, so does the GPU compute wasted recalculating work a model has already done. Augmented Memory Grid, a NeuralMesh 6 feature built specifically for this, is Weka’s answer.Every prompt triggers two stages. Prefill calculates attention, the core mechanism behind how large language models process input, and it’s computationally expensive. Decode converts that calculation into output and is comparatively lightweight. The cost shows up hardest in multi-turn sessions like chat or coding, where each new turn re-triggers prefill for everything that came before it, unless that work has been cached.”If you have 10 turns, you may overcalculate 100 times because you’re redoing all of them. If you have 20, you’ll overcalculate 400 times,” Zvibel said. “You can put two orders of magnitude more NAND than you could afford in shared memory, and we can cache 100% of the pre-calculated tokens, so you never need to redo it.”Where Weka sits competitivelyStorage vendors have spent the past year and a half repositioning around AI, and separating genuine capability from repositioned messaging is now a real evaluation problem for buyers. “The storage world is shifting its focus from serving bits to enterprise workloads to managing data at the speed of AI. We’ve seen that most clearly over the past 18 months from Dell, NetApp, and Pure,” Steve McDowell, chief analyst at NAND Research, told VentureBeat. “The interesting thing is that companies like Weka, and VAST, are the true AI-native data companies, solving these problems since day one.”McDowell singled out Augmented Memory Grid as Weka’s clearest technical lead. “Weka continues to have the most technically capable KV cache implementation on the market with its Augmented Memory Grid,” he said. ” They were early with this technology, and continue to innovate. This is critical for AI inference, as it enables a level of GPU efficiency that, without question, saves money on GPUs and memory. That’s key for today’s memory and GPU constrained market.” He also flagged Weka’s contractual guarantee on its data reduction claims as underappreciated. “One flying a little under the radar: Weka is putting its money where its mouth is with its contractual guarantees for its data reduction promises,” he said.McDowell’s advice to buyers evaluating competing claims from Weka, VAST, Pure and NetApp alike was pointed suggesting that enterprise buyers should look hard at what vendors are promising versus what they’re actually delivering.”A smart buyer will look at how competing vendors are solving real-world problems today,” McDowell said. ” They do this by talking to organizations running similar workloads at similar scale. If a vendor can’t point to that, then it should be a warning sign.”

When I first stepped into leadership, I believed something that many ambitious professionals believe: Leadership meant having all the answers. I thought my responsibility as a leader was to be the person everyone turned to for solutions. If a problem surfaced, I needed to solve it. If uncertainty emerged, I needed to eliminate it. If […]

One developer intends to quit his job at JPMorgan Chase next month to build apps full-time.